X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/490019cf9519204c5fb36b2fba54ceb983bb6b72..0a7de7458d150b5d4dffc935ba399be265ef0a1a:/bsd/kern/kern_exec.c diff --git a/bsd/kern/kern_exec.c b/bsd/kern/kern_exec.c index 20b1f0317..fea51a172 100644 --- a/bsd/kern/kern_exec.c +++ b/bsd/kern/kern_exec.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000-2011 Apple Inc. All rights reserved. + * Copyright (c) 2000-2013 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ - * + * * This file contains Original Code and/or Modifications of Original Code * as defined in and that are subject to the Apple Public Source License * Version 2.0 (the 'License'). You may not use this file except in @@ -11,10 +11,10 @@ * unlawful or unlicensed copies of an Apple operating system, or to * circumvent, violate, or enable the circumvention or violation of, any * terms of an Apple operating system software license agreement. - * + * * Please obtain a copy of the License at * http://www.opensource.apple.com/apsl/ and read it before using this file. - * + * * The Original Code and all software distributed under the License are * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, @@ -22,7 +22,7 @@ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. * Please see the License for the specific language governing rights and * limitations under the License. - * + * * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */ @@ -32,7 +32,7 @@ * All rights reserved. The CMU software License Agreement specifies * the terms and conditions for use and redistribution. */ - + /*- * Copyright (c) 1982, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. @@ -92,7 +92,7 @@ #include #include #include -#include +#include #include #include #include @@ -103,14 +103,16 @@ #include #include #include +#include #if SYSV_SHM -#include /* shmexec() */ +#include /* shmexec() */ #endif -#include /* ubc_map() */ +#include /* ubc_map() */ #include #include #include #include +#include #include #include @@ -133,10 +135,13 @@ #include #include #include +#include #include +#include + #if CONFIG_MACF -#include +#include #include #endif @@ -157,12 +162,14 @@ #include #endif +#include + +extern boolean_t vm_darkwake_mode; + #if CONFIG_DTRACE /* Do not include dtrace.h, it redefines kmem_[alloc/free] */ -extern void (*dtrace_fasttrap_exec_ptr)(proc_t); +extern void dtrace_proc_exec(proc_t); extern void (*dtrace_proc_waitfor_exec_ptr)(proc_t); -extern void (*dtrace_helpers_cleanup)(proc_t); -extern void dtrace_lazy_dofs_destroy(proc_t); /* * Since dtrace_proc_waitfor_exec_ptr can be added/removed in dtrace_subr.c, @@ -174,22 +181,36 @@ static void (*dtrace_proc_waitfor_hook)(proc_t) = NULL; #endif /* support for child creation in exec after vfork */ -thread_t fork_create_child(task_t parent_task, coalition_t *parent_coalition, proc_t child_proc, int inherit_memory, int is64bit); +thread_t fork_create_child(task_t parent_task, + coalition_t *parent_coalition, + proc_t child_proc, + int inherit_memory, + int is_64bit_addr, + int is_64bit_data, + int in_exec); void vfork_exit(proc_t p, int rv); extern void proc_apply_task_networkbg_internal(proc_t, thread_t); +extern void task_set_did_exec_flag(task_t task); +extern void task_clear_exec_copy_flag(task_t task); +proc_t proc_exec_switch_task(proc_t p, task_t old_task, task_t new_task, thread_t new_thread); +boolean_t task_is_active(task_t); +boolean_t thread_is_active(thread_t thread); +void thread_copy_resource_info(thread_t dst_thread, thread_t src_thread); +void *ipc_importance_exec_switch_task(task_t old_task, task_t new_task); +extern void ipc_importance_release(void *elem); /* * Mach things for which prototypes are unavailable from Mach headers */ -void ipc_task_reset( - task_t task); -void ipc_thread_reset( - thread_t thread); +void ipc_task_reset( + task_t task); +void ipc_thread_reset( + thread_t thread); kern_return_t ipc_object_copyin( - ipc_space_t space, - mach_port_name_t name, - mach_msg_type_name_t msgt_name, - ipc_object_t *objectp); + ipc_space_t space, + mach_port_name_t name, + mach_msg_type_name_t msgt_name, + ipc_object_t *objectp); void ipc_port_release_send(ipc_port_t); #if DEVELOPMENT || DEBUG @@ -197,7 +218,6 @@ void task_importance_update_owner_info(task_t); #endif extern struct savearea *get_user_regs(thread_t); -extern kern_return_t machine_thread_neon_state_initialize(thread_t thread); __attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__(mach_port_t task_access_port, int32_t new_pid); @@ -219,7 +239,7 @@ __attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL_ * activator in exec_activate_image() before treating * it as malformed/corrupt. */ -#define EAI_ITERLIMIT 3 +#define EAI_ITERLIMIT 3 /* * For #! interpreter parsing @@ -229,12 +249,23 @@ __attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL_ extern vm_map_t bsd_pageable_map; extern const struct fileops vnops; +extern int nextpidversion; -#define USER_ADDR_ALIGN(addr, val) \ +#define USER_ADDR_ALIGN(addr, val) \ ( ( (user_addr_t)(addr) + (val) - 1) \ - & ~((val) - 1) ) + & ~((val) - 1) ) -struct image_params; /* Forward */ +/* Platform Code Exec Logging */ +static int platform_exec_logging = 0; + +SYSCTL_DECL(_security_mac); + +SYSCTL_INT(_security_mac, OID_AUTO, platform_exec_logging, CTLFLAG_RW, &platform_exec_logging, 0, + "log cdhashes for all platform binary executions"); + +static os_log_t peLog = OS_LOG_DEFAULT; + +struct image_params; /* Forward */ static int exec_activate_image(struct image_params *imgp); static int exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp); static int load_return_to_errno(load_return_t lrtn); @@ -242,18 +273,18 @@ static int execargs_alloc(struct image_params *imgp); static int execargs_free(struct image_params *imgp); static int exec_check_permissions(struct image_params *imgp); static int exec_extract_strings(struct image_params *imgp); -static int exec_add_apple_strings(struct image_params *imgp); +static int exec_add_apple_strings(struct image_params *imgp, const load_result_t *load_result); static int exec_handle_sugid(struct image_params *imgp); static int sugid_scripts = 0; -SYSCTL_INT (_kern, OID_AUTO, sugid_scripts, CTLFLAG_RW | CTLFLAG_LOCKED, &sugid_scripts, 0, ""); +SYSCTL_INT(_kern, OID_AUTO, sugid_scripts, CTLFLAG_RW | CTLFLAG_LOCKED, &sugid_scripts, 0, ""); static kern_return_t create_unix_stack(vm_map_t map, load_result_t* load_result, proc_t p); static int copyoutptr(user_addr_t ua, user_addr_t ptr, int ptr_size); static void exec_resettextvp(proc_t, struct image_params *); static int check_for_signature(proc_t, struct image_params *); static void exec_prefault_data(proc_t, struct image_params *, load_result_t *); -static errno_t exec_handle_port_actions(struct image_params *imgp, short psa_flags, boolean_t * portwatch_present, ipc_port_t * portwatch_ports); +static errno_t exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_present, ipc_port_t * portwatch_ports); static errno_t exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, uint64_t psa_darwin_role, - ipc_port_t * portwatch_ports, int portwatch_count); + ipc_port_t * portwatch_ports, int portwatch_count); /* * exec_add_user_string @@ -277,35 +308,36 @@ static int exec_add_user_string(struct image_params *imgp, user_addr_t str, int seg, boolean_t is_ncargs) { int error = 0; - + do { size_t len = 0; int space; - - if (is_ncargs) + + if (is_ncargs) { space = imgp->ip_argspace; /* by definition smaller than ip_strspace */ - else + } else { space = imgp->ip_strspace; - + } + if (space <= 0) { error = E2BIG; break; } - + if (!UIO_SEG_IS_USER_SPACE(seg)) { - char *kstr = CAST_DOWN(char *,str); /* SAFE */ + char *kstr = CAST_DOWN(char *, str); /* SAFE */ error = copystr(kstr, imgp->ip_strendp, space, &len); - } else { + } else { error = copyinstr(str, imgp->ip_strendp, space, &len); } imgp->ip_strendp += len; imgp->ip_strspace -= len; - if (is_ncargs) + if (is_ncargs) { imgp->ip_argspace -= len; - + } } while (error == ENAMETOOLONG); - + return error; } @@ -313,7 +345,7 @@ exec_add_user_string(struct image_params *imgp, user_addr_t str, int seg, boolea * dyld is now passed the executable path as a getenv-like variable * in the same fashion as the stack_guard and malloc_entropy keys. */ -#define EXECUTABLE_KEY "executable_path=" +#define EXECUTABLE_KEY "executable_path=" /* * exec_save_path @@ -362,13 +394,13 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char len = MIN(MAXPATHLEN, imgp->ip_strspace); - switch(seg) { + switch (seg) { case UIO_USERSPACE32: - case UIO_USERSPACE64: /* Same for copyin()... */ + case UIO_USERSPACE64: /* Same for copyin()... */ error = copyinstr(path, imgp->ip_strings + strlen(EXECUTABLE_KEY), len, &len); break; case UIO_SYSSPACE: - kpath = CAST_DOWN(char *,path); /* SAFE */ + kpath = CAST_DOWN(char *, path); /* SAFE */ error = copystr(kpath, imgp->ip_strings + strlen(EXECUTABLE_KEY), len, &len); break; default: @@ -388,7 +420,7 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char } } - return(error); + return error; } /* @@ -396,7 +428,7 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char * * If we detect a shell script, we need to reset the string area * state so that the interpreter can be saved onto the stack. - + * * Parameters; struct image_params * image parameter block * * Returns: int 0 Success @@ -413,9 +445,9 @@ exec_reset_save_path(struct image_params *imgp) { imgp->ip_strendp = imgp->ip_strings; imgp->ip_argspace = NCARGS; - imgp->ip_strspace = ( NCARGS + PAGE_SIZE ); + imgp->ip_strspace = (NCARGS + PAGE_SIZE); - return (0); + return 0; } /* @@ -446,10 +478,6 @@ exec_shell_imgact(struct image_params *imgp) char *ihp; char *line_startp, *line_endp; char *interp; - proc_t p; - struct fileproc *fp; - int fd; - int error; /* * Make sure it's a shell script. If we've already redirected @@ -458,12 +486,12 @@ exec_shell_imgact(struct image_params *imgp) if (vdata[0] != '#' || vdata[1] != '!' || (imgp->ip_flags & IMGPF_INTERPRET) != 0) { - return (-1); + return -1; } if (imgp->ip_origcputype != 0) { /* Fat header previously matched, don't allow shell script inside */ - return (-1); + return -1; } imgp->ip_flags |= IMGPF_INTERPRET; @@ -479,10 +507,10 @@ exec_shell_imgact(struct image_params *imgp) } /* Try to find the first non-whitespace character */ - for( ihp = &vdata[2]; ihp < &vdata[IMG_SHSIZE]; ihp++ ) { + for (ihp = &vdata[2]; ihp < &vdata[IMG_SHSIZE]; ihp++) { if (IS_EOL(*ihp)) { /* Did not find interpreter, "#!\n" */ - return (ENOEXEC); + return ENOEXEC; } else if (IS_WHITESPACE(*ihp)) { /* Whitespace, like "#! /bin/sh\n", keep going. */ } else { @@ -493,13 +521,13 @@ exec_shell_imgact(struct image_params *imgp) if (ihp == &vdata[IMG_SHSIZE]) { /* All whitespace, like "#! " */ - return (ENOEXEC); + return ENOEXEC; } line_startp = ihp; /* Try to find the end of the interpreter+args string */ - for ( ; ihp < &vdata[IMG_SHSIZE]; ihp++ ) { + for (; ihp < &vdata[IMG_SHSIZE]; ihp++) { if (IS_EOL(*ihp)) { /* Got it */ break; @@ -510,7 +538,7 @@ exec_shell_imgact(struct image_params *imgp) if (ihp == &vdata[IMG_SHSIZE]) { /* A long line, like "#! blah blah blah" without end */ - return (ENOEXEC); + return ENOEXEC; } /* Backtrack until we find the last non-whitespace */ @@ -530,35 +558,44 @@ exec_shell_imgact(struct image_params *imgp) /* copy the interpreter name */ interp = imgp->ip_interp_buffer; - for ( ihp = line_startp; (ihp < line_endp) && !IS_WHITESPACE(*ihp); ihp++) + for (ihp = line_startp; (ihp < line_endp) && !IS_WHITESPACE(*ihp); ihp++) { *interp++ = *ihp; + } *interp = '\0'; exec_reset_save_path(imgp); exec_save_path(imgp, CAST_USER_ADDR_T(imgp->ip_interp_buffer), - UIO_SYSSPACE, NULL); + UIO_SYSSPACE, NULL); /* Copy the entire interpreter + args for later processing into argv[] */ interp = imgp->ip_interp_buffer; - for ( ihp = line_startp; (ihp < line_endp); ihp++) + for (ihp = line_startp; (ihp < line_endp); ihp++) { *interp++ = *ihp; + } *interp = '\0'; +#if !SECURE_KERNEL /* - * If we have a SUID oder SGID script, create a file descriptor + * If we have an SUID or SGID script, create a file descriptor * from the vnode and pass /dev/fd/%d instead of the actual * path name so that the script does not get opened twice */ if (imgp->ip_origvattr->va_mode & (VSUID | VSGID)) { + proc_t p; + struct fileproc *fp; + int fd; + int error; + p = vfs_context_proc(imgp->ip_vfs_context); error = falloc(p, &fp, &fd, imgp->ip_vfs_context); - if (error) - return(error); + if (error) { + return error; + } fp->f_fglob->fg_flag = FREAD; fp->f_fglob->fg_ops = &vnops; fp->f_fglob->fg_data = (caddr_t)imgp->ip_vp; - + proc_fdlock(p); procfdtbl_releasefd(p, fd, NULL); fp_drop(p, fd, fp, 1); @@ -567,8 +604,9 @@ exec_shell_imgact(struct image_params *imgp) imgp->ip_interp_sugid_fd = fd; } +#endif - return (-3); + return -3; } @@ -594,7 +632,7 @@ exec_shell_imgact(struct image_params *imgp) * activators should not be given the opportunity to attempt * to activate the image. * - * If we find an encapsulated binary, we make no assertions + * If we find an encapsulated binary, we make no assertions * about its validity; instead, we leave that up to a rescan * for an activator to claim it, and, if it is claimed by one, * that activator is responsible for determining validity. @@ -612,7 +650,8 @@ exec_fat_imgact(struct image_params *imgp) if (imgp->ip_origcputype != 0) { /* Fat header previously matched, don't allow another fat file inside */ - return (-1); + error = -1; /* not claimed */ + goto bad; } /* Make sure it's a fat binary */ @@ -648,9 +687,9 @@ exec_fat_imgact(struct image_params *imgp) } lret = fatfile_getbestarch_for_cputype(pref, - (vm_offset_t)fat_header, - PAGE_SIZE, - &fat_arch); + (vm_offset_t)fat_header, + PAGE_SIZE, + &fat_arch); if (lret == LOAD_SUCCESS) { goto use_arch; } @@ -664,8 +703,8 @@ exec_fat_imgact(struct image_params *imgp) regular_grading: /* Look up our preferred architecture in the fat file. */ lret = fatfile_getbestarch((vm_offset_t)fat_header, - PAGE_SIZE, - &fat_arch); + PAGE_SIZE, + &fat_arch); if (lret != LOAD_SUCCESS) { error = load_return_to_errno(lret); goto bad; @@ -674,9 +713,9 @@ regular_grading: use_arch: /* Read the Mach-O header out of fat_arch */ error = vn_rdwr(UIO_READ, imgp->ip_vp, imgp->ip_vdata, - PAGE_SIZE, fat_arch.offset, - UIO_SYSSPACE, (IO_UNIT|IO_NODELOCKED), - cred, &resid, p); + PAGE_SIZE, fat_arch.offset, + UIO_SYSSPACE, (IO_UNIT | IO_NODELOCKED), + cred, &resid, p); if (error) { goto bad; } @@ -694,20 +733,27 @@ use_arch: bad: kauth_cred_unref(&cred); - return (error); + return error; } static int -activate_thread_state(thread_t thread, load_result_t *result) +activate_exec_state(task_t task, proc_t p, thread_t thread, load_result_t *result) { int ret; + task_set_dyld_info(task, MACH_VM_MIN_ADDRESS, 0); + task_set_64bit(task, result->is_64bit_addr, result->is_64bit_data); + if (result->is_64bit_addr) { + OSBitOrAtomic(P_LP64, &p->p_flag); + } else { + OSBitAndAtomic(~((uint32_t)P_LP64), &p->p_flag); + } + ret = thread_state_initialize(thread); if (ret != KERN_SUCCESS) { return ret; } - if (result->threadstate) { uint32_t *ts = result->threadstate; uint32_t total_size = result->threadstate_sz; @@ -731,6 +777,31 @@ activate_thread_state(thread_t thread, load_result_t *result) } +/* + * Set p->p_comm and p->p_name to the name passed to exec + */ +static void +set_proc_name(struct image_params *imgp, proc_t p) +{ + int p_name_len = sizeof(p->p_name) - 1; + + if (imgp->ip_ndp->ni_cnd.cn_namelen > p_name_len) { + imgp->ip_ndp->ni_cnd.cn_namelen = p_name_len; + } + + bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_name, + (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); + p->p_name[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; + + if (imgp->ip_ndp->ni_cnd.cn_namelen > MAXCOMLEN) { + imgp->ip_ndp->ni_cnd.cn_namelen = MAXCOMLEN; + } + + bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_comm, + (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); + p->p_comm[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; +} + /* * exec_mach_imgact * @@ -758,27 +829,28 @@ static int exec_mach_imgact(struct image_params *imgp) { struct mach_header *mach_header = (struct mach_header *)imgp->ip_vdata; - proc_t p = vfs_context_proc(imgp->ip_vfs_context); - int error = 0; - task_t task; - task_t new_task = NULL; /* protected by vfexec */ - thread_t thread; - struct uthread *uthread; + proc_t p = vfs_context_proc(imgp->ip_vfs_context); + int error = 0; + task_t task; + task_t new_task = NULL; /* protected by vfexec */ + thread_t thread; + struct uthread *uthread; vm_map_t old_map = VM_MAP_NULL; - vm_map_t map; - load_return_t lret; - load_result_t load_result; + vm_map_t map = VM_MAP_NULL; + load_return_t lret; + load_result_t load_result = {}; struct _posix_spawnattr *psa = NULL; - int spawn = (imgp->ip_flags & IMGPF_SPAWN); - int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); - int p_name_len; + int spawn = (imgp->ip_flags & IMGPF_SPAWN); + int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); + int exec = (imgp->ip_flags & IMGPF_EXEC); + os_reason_t exec_failure_reason = OS_REASON_NULL; /* * make sure it's a Mach-O 1.0 or Mach-O 2.0 binary; the difference * is a reserved field on the end, so for the most part, we can * treat them as if they were identical. Reverse-endian Mach-O * binaries are recognized but not compatible. - */ + */ if ((mach_header->magic == MH_CIGAM) || (mach_header->magic == MH_CIGAM_64)) { error = EBADARCH; @@ -799,7 +871,7 @@ exec_mach_imgact(struct image_params *imgp) if (imgp->ip_origcputype != 0) { /* Fat header previously had an idea about this thin file */ if (imgp->ip_origcputype != mach_header->cputype || - imgp->ip_origcpusubtype != mach_header->cpusubtype) { + imgp->ip_origcpusubtype != mach_header->cpusubtype) { error = EBADARCH; goto bad; } @@ -812,8 +884,9 @@ exec_mach_imgact(struct image_params *imgp) thread = current_thread(); uthread = get_bsdthread_info(thread); - if ((mach_header->cputype & CPU_ARCH_ABI64) == CPU_ARCH_ABI64) - imgp->ip_flags |= IMGPF_IS_64BIT; + if ((mach_header->cputype & CPU_ARCH_ABI64) == CPU_ARCH_ABI64) { + imgp->ip_flags |= IMGPF_IS_64BIT_ADDR | IMGPF_IS_64BIT_DATA; + } /* If posix_spawn binprefs exist, respect those prefs. */ psa = (struct _posix_spawnattr *) imgp->ip_px_sa; @@ -846,16 +919,14 @@ grade: goto bad; } + /* Copy in arguments/environment from the old process */ error = exec_extract_strings(imgp); - if (error) - goto bad; - - error = exec_add_apple_strings(imgp); - if (error) + if (error) { goto bad; + } - AUDIT_ARG(argv, imgp->ip_startargv, imgp->ip_argc, + AUDIT_ARG(argv, imgp->ip_startargv, imgp->ip_argc, imgp->ip_endargv - imgp->ip_startargv); AUDIT_ARG(envv, imgp->ip_endargv, imgp->ip_envc, imgp->ip_endenvv - imgp->ip_endargv); @@ -867,36 +938,26 @@ grade: * obtained indirectly from the image_params vfs_context_t, is the * new child process. */ - if (vfexec || spawn) { - if (vfexec) { - imgp->ip_new_thread = fork_create_child(task, NULL, p, FALSE, (imgp->ip_flags & IMGPF_IS_64BIT)); - if (imgp->ip_new_thread == NULL) { - error = ENOMEM; - goto bad; - } + if (vfexec) { + imgp->ip_new_thread = fork_create_child(task, + NULL, + p, + FALSE, + (imgp->ip_flags & IMGPF_IS_64BIT_ADDR), + (imgp->ip_flags & IMGPF_IS_64BIT_DATA), + FALSE); + /* task and thread ref returned, will be released in __mac_execve */ + if (imgp->ip_new_thread == NULL) { + error = ENOMEM; + goto bad; } - - /* reset local idea of thread, uthread, task */ - thread = imgp->ip_new_thread; - uthread = get_bsdthread_info(thread); - task = new_task = get_threadtask(thread); - map = get_task_map(task); - } else { - map = VM_MAP_NULL; } - /* - * We set these flags here; this is OK, since if we fail after - * this point, we have already destroyed the parent process anyway. - */ - task_set_dyld_info(task, MACH_VM_MIN_ADDRESS, 0); - if (imgp->ip_flags & IMGPF_IS_64BIT) { - task_set_64bit(task, TRUE); - OSBitOrAtomic(P_LP64, &p->p_flag); - } else { - task_set_64bit(task, FALSE); - OSBitAndAtomic(~((uint32_t)P_LP64), &p->p_flag); - } + + /* reset local idea of thread, uthread, task */ + thread = imgp->ip_new_thread; + uthread = get_bsdthread_info(thread); + task = new_task = get_threadtask(thread); /* * Load the Mach-O file. @@ -914,9 +975,29 @@ grade: * Actually load the image file we previously decided to load. */ lret = load_machfile(imgp, mach_header, thread, &map, &load_result); - if (lret != LOAD_SUCCESS) { error = load_return_to_errno(lret); + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_BAD_MACHO, 0, 0); + if (lret == LOAD_BADMACHO_UPX) { + /* set anything that might be useful in the crash report */ + set_proc_name(imgp, p); + + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_UPX); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + exec_failure_reason->osr_flags |= OS_REASON_FLAG_CONSISTENT_FAILURE; + } else if (lret == LOAD_BADARCH_X86) { + /* set anything that might be useful in the crash report */ + set_proc_name(imgp, p); + + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_NO32EXEC); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + exec_failure_reason->osr_flags |= OS_REASON_FLAG_CONSISTENT_FAILURE; + } else { + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_BAD_MACHO); + } + goto badtoolate; } @@ -927,90 +1008,128 @@ grade: vm_map_set_user_wire_limit(map, p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur); - /* + /* * Set code-signing flags if this binary is signed, or if parent has * requested them on exec. */ if (load_result.csflags & CS_VALID) { - imgp->ip_csflags |= load_result.csflags & - (CS_VALID| - CS_HARD|CS_KILL|CS_RESTRICT|CS_ENFORCEMENT|CS_REQUIRE_LV|CS_DYLD_PLATFORM| - CS_EXEC_SET_HARD|CS_EXEC_SET_KILL|CS_EXEC_SET_ENFORCEMENT); + imgp->ip_csflags |= load_result.csflags & + (CS_VALID | CS_SIGNED | CS_DEV_CODE | + CS_HARD | CS_KILL | CS_RESTRICT | CS_ENFORCEMENT | CS_REQUIRE_LV | + CS_FORCED_LV | CS_ENTITLEMENTS_VALIDATED | CS_DYLD_PLATFORM | CS_RUNTIME | + CS_ENTITLEMENT_FLAGS | + CS_EXEC_SET_HARD | CS_EXEC_SET_KILL | CS_EXEC_SET_ENFORCEMENT); } else { imgp->ip_csflags &= ~CS_VALID; } - if (p->p_csflags & CS_EXEC_SET_HARD) + if (p->p_csflags & CS_EXEC_SET_HARD) { imgp->ip_csflags |= CS_HARD; - if (p->p_csflags & CS_EXEC_SET_KILL) + } + if (p->p_csflags & CS_EXEC_SET_KILL) { imgp->ip_csflags |= CS_KILL; - if (p->p_csflags & CS_EXEC_SET_ENFORCEMENT) + } + if (p->p_csflags & CS_EXEC_SET_ENFORCEMENT) { imgp->ip_csflags |= CS_ENFORCEMENT; - if (p->p_csflags & CS_EXEC_SET_INSTALLER) - imgp->ip_csflags |= CS_INSTALLER; + } + if (p->p_csflags & CS_EXEC_INHERIT_SIP) { + if (p->p_csflags & CS_INSTALLER) { + imgp->ip_csflags |= CS_INSTALLER; + } + if (p->p_csflags & CS_DATAVAULT_CONTROLLER) { + imgp->ip_csflags |= CS_DATAVAULT_CONTROLLER; + } + if (p->p_csflags & CS_NVRAM_UNRESTRICTED) { + imgp->ip_csflags |= CS_NVRAM_UNRESTRICTED; + } + } /* * Set up the system reserved areas in the new address space. */ - vm_map_exec(map, task, (void *)p->p_fd->fd_rdir, cpu_type()); + int cpu_subtype; + cpu_subtype = 0; /* all cpu_subtypes use the same shared region */ + vm_map_exec(map, task, load_result.is_64bit_addr, (void *)p->p_fd->fd_rdir, cpu_type(), cpu_subtype); /* * Close file descriptors which specify close-on-exec. */ - fdexec(p, psa != NULL ? psa->psa_flags : 0); + fdexec(p, psa != NULL ? psa->psa_flags : 0, exec); /* * deal with set[ug]id. */ error = exec_handle_sugid(imgp); if (error) { - if (spawn || !vfexec) { - vm_map_deallocate(map); - } + vm_map_deallocate(map); + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SUGID_FAILURE, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_SUGID_FAILURE); goto badtoolate; } /* * Commit to new map. * - * Swap the new map for the old, which consumes our new map reference but - * each leaves us responsible for the old_map reference. That lets us get - * off the pmap associated with it, and then we can release it. + * Swap the new map for the old for target task, which consumes + * our new map reference but each leaves us responsible for the + * old_map reference. That lets us get off the pmap associated + * with it, and then we can release it. + * + * The map needs to be set on the target task which is different + * than current task, thus swap_task_map is used instead of + * vm_map_switch. */ - if (!vfexec) { - old_map = swap_task_map(task, thread, map, !spawn); - vm_map_deallocate(old_map); - } + old_map = swap_task_map(task, thread, map); + vm_map_deallocate(old_map); + old_map = NULL; - lret = activate_thread_state(thread, &load_result); + lret = activate_exec_state(task, p, thread, &load_result); if (lret != KERN_SUCCESS) { + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_ACTV_THREADSTATE, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_ACTV_THREADSTATE); goto badtoolate; } /* * deal with voucher on exec-calling thread. */ - if (imgp->ip_new_thread == NULL) + if (imgp->ip_new_thread == NULL) { thread_set_mach_voucher(current_thread(), IPC_VOUCHER_NULL); + } /* Make sure we won't interrupt ourself signalling a partial process */ - if (!vfexec && !spawn && (p->p_lflag & P_LTRACED)) + if (!vfexec && !spawn && (p->p_lflag & P_LTRACED)) { psignal(p, SIGTRAP); + } if (load_result.unixproc && - create_unix_stack(get_task_map(task), - &load_result, - p) != KERN_SUCCESS) { + create_unix_stack(get_task_map(task), + &load_result, + p) != KERN_SUCCESS) { error = load_return_to_errno(LOAD_NOSPACE); + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_STACK_ALLOC, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_STACK_ALLOC); goto badtoolate; } - if (vfexec || spawn) { - old_map = vm_map_switch(get_task_map(task)); + error = exec_add_apple_strings(imgp, &load_result); + if (error) { + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_APPLE_STRING_INIT, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_APPLE_STRING_INIT); + goto badtoolate; } + /* Switch to target task's map to copy out strings */ + old_map = vm_map_switch(get_task_map(task)); + if (load_result.unixproc) { - user_addr_t ap; + user_addr_t ap; /* * Copy the strings area out into the new process address @@ -1019,8 +1138,11 @@ grade: ap = p->user_stack; error = exec_copyout_strings(imgp, &ap); if (error) { - if (vfexec || spawn) - vm_map_switch(old_map); + vm_map_switch(old_map); + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_STRINGS, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_STRINGS); goto badtoolate; } /* Set the stack */ @@ -1028,16 +1150,19 @@ grade: } if (load_result.dynlinker) { - uint64_t ap; - int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; + uint64_t ap; + int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; /* Adjust the stack */ ap = thread_adjuserstack(thread, -new_ptr_size); error = copyoutptr(load_result.mach_header, ap, new_ptr_size); if (error) { - if (vfexec || spawn) - vm_map_switch(old_map); + vm_map_switch(old_map); + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_DYNLINKER, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_DYNLINKER); goto badtoolate; } task_set_dyld_info(task, load_result.all_image_info_addr, @@ -1047,9 +1172,7 @@ grade: /* Avoid immediate VM faults back into kernel */ exec_prefault_data(p, imgp, &load_result); - if (vfexec || spawn) { - vm_map_switch(old_map); - } + vm_map_switch(old_map); /* Stop profiling */ stopprofclock(p); @@ -1067,8 +1190,9 @@ grade: #if SYSV_SHM /* FIXME: Till vmspace inherit is fixed: */ - if (!vfexec && p->vm_shm) + if (!vfexec && p->vm_shm) { shmexec(p); + } #endif #if SYSV_SEM /* Clean up the semaphores */ @@ -1080,89 +1204,97 @@ grade: */ p->p_acflag &= ~AFORK; + set_proc_name(imgp, p); + +#if CONFIG_SECLUDED_MEMORY + if (secluded_for_apps && + load_result.platform_binary) { + if (strncmp(p->p_name, + "Camera", + sizeof(p->p_name)) == 0) { + task_set_could_use_secluded_mem(task, TRUE); + } else { + task_set_could_use_secluded_mem(task, FALSE); + } + if (strncmp(p->p_name, + "mediaserverd", + sizeof(p->p_name)) == 0) { + task_set_could_also_use_secluded_mem(task, TRUE); + } + } +#endif /* CONFIG_SECLUDED_MEMORY */ + +#if __arm64__ + if (load_result.legacy_footprint) { +#if DEVELOPMENT || DEBUG + printf("%s: %d[%s] legacy footprint (mach-o)\n", + __FUNCTION__, p->p_pid, p->p_name); +#endif /* DEVELOPMENT || DEBUG */ + task_set_legacy_footprint(task, TRUE); + } +#endif /* __arm64__ */ + + pal_dbg_set_task_name(task); + /* - * Set p->p_comm and p->p_name to the name passed to exec + * The load result will have already been munged by AMFI to include the + * platform binary flag if boot-args dictated it (AMFI will mark anything + * that doesn't go through the upcall path as a platform binary if its + * enforcement is disabled). */ - p_name_len = sizeof(p->p_name) - 1; - if(imgp->ip_ndp->ni_cnd.cn_namelen > p_name_len) - imgp->ip_ndp->ni_cnd.cn_namelen = p_name_len; - bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_name, - (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); - p->p_name[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; + if (load_result.platform_binary) { + if (cs_debug) { + printf("setting platform binary on task: pid = %d\n", p->p_pid); + } - if (imgp->ip_ndp->ni_cnd.cn_namelen > MAXCOMLEN) - imgp->ip_ndp->ni_cnd.cn_namelen = MAXCOMLEN; - bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_comm, - (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); - p->p_comm[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; + /* + * We must use 'task' here because the proc's task has not yet been + * switched to the new one. + */ + task_set_platform_binary(task, TRUE); + } else { + if (cs_debug) { + printf("clearing platform binary on task: pid = %d\n", p->p_pid); + } - pal_dbg_set_task_name( p->task ); + task_set_platform_binary(task, FALSE); + } #if DEVELOPMENT || DEBUG - /* + /* * Update the pid an proc name for importance base if any */ - task_importance_update_owner_info(p->task); + task_importance_update_owner_info(task); #endif memcpy(&p->p_uuid[0], &load_result.uuid[0], sizeof(p->p_uuid)); -// dtrace code cleanup needed #if CONFIG_DTRACE - /* - * Invalidate any predicate evaluation already cached for this thread by DTrace. - * That's because we've just stored to p_comm and DTrace refers to that when it - * evaluates the "execname" special variable. uid and gid may have changed as well. - */ - dtrace_set_thread_predcache(current_thread(), 0); - - /* - * Free any outstanding lazy dof entries. It is imperative we - * always call dtrace_lazy_dofs_destroy, rather than null check - * and call if !NULL. If we NULL test, during lazy dof faulting - * we can race with the faulting code and proceed from here to - * beyond the helpers cleanup. The lazy dof faulting will then - * install new helpers which no longer belong to this process! - */ - dtrace_lazy_dofs_destroy(p); - - - /* - * Clean up any DTrace helpers for the process. - */ - if (p->p_dtrace_helpers != NULL && dtrace_helpers_cleanup) { - (*dtrace_helpers_cleanup)(p); - } - - /* - * Cleanup the DTrace provider associated with this process. - */ - proc_lock(p); - if (p->p_dtrace_probes && dtrace_fasttrap_exec_ptr) { - (*dtrace_fasttrap_exec_ptr)(p); - } - proc_unlock(p); + dtrace_proc_exec(p); #endif if (kdebug_enable) { - long dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4; + long args[4] = {}; + + uintptr_t fsid = 0, fileid = 0; + if (imgp->ip_vattr) { + uint64_t fsid64 = vnode_get_va_fsid(imgp->ip_vattr); + fsid = fsid64; + fileid = imgp->ip_vattr->va_fileid; + // check for (unexpected) overflow and trace zero in that case + if (fsid != fsid64 || fileid != imgp->ip_vattr->va_fileid) { + fsid = fileid = 0; + } + } + KERNEL_DEBUG_CONSTANT_IST1(TRACE_DATA_EXEC, p->p_pid, fsid, fileid, 0, + (uintptr_t)thread_tid(thread)); /* * Collect the pathname for tracing */ - kdbg_trace_string(p, &dbg_arg1, &dbg_arg2, &dbg_arg3, &dbg_arg4); - - if (vfexec || spawn) { - KERNEL_DEBUG_CONSTANT1(TRACE_DATA_EXEC | DBG_FUNC_NONE, - p->p_pid ,0,0,0, (uintptr_t)thread_tid(thread)); - KERNEL_DEBUG_CONSTANT1(TRACE_STRING_EXEC | DBG_FUNC_NONE, - dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4, (uintptr_t)thread_tid(thread)); - } else { - KERNEL_DEBUG_CONSTANT(TRACE_DATA_EXEC | DBG_FUNC_NONE, - p->p_pid ,0,0,0,0); - KERNEL_DEBUG_CONSTANT(TRACE_STRING_EXEC | DBG_FUNC_NONE, - dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4, 0); - } + kdbg_trace_string(p, &args[0], &args[1], &args[2], &args[3]); + KERNEL_DEBUG_CONSTANT_IST1(TRACE_STRING_EXEC, args[0], args[1], + args[2], args[3], (uintptr_t)thread_tid(thread)); } /* @@ -1175,7 +1307,7 @@ grade: proc_lock(p); p->p_stat = SSTOP; proc_unlock(p); - (void) task_suspend_internal(p->task); + (void) task_suspend_internal(task); } } @@ -1206,35 +1338,37 @@ badtoolate: /* Don't allow child process to execute any instructions */ if (!spawn) { if (vfexec) { - psignal_vfork(p, new_task, thread, SIGKILL); + assert(exec_failure_reason != OS_REASON_NULL); + psignal_vfork_with_reason(p, new_task, thread, SIGKILL, exec_failure_reason); + exec_failure_reason = OS_REASON_NULL; } else { - psignal(p, SIGKILL); + assert(exec_failure_reason != OS_REASON_NULL); + psignal_with_reason(p, SIGKILL, exec_failure_reason); + exec_failure_reason = OS_REASON_NULL; + + if (exec) { + /* Terminate the exec copy task */ + task_terminate_internal(task); + } } /* We can't stop this system call at this point, so just pretend we succeeded */ error = 0; - } - -done: - if (!spawn) { - /* notify only if it has not failed due to FP Key error */ - if ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0) - proc_knote(p, NOTE_EXEC); - } - - /* Drop extra references for cases where we don't expect the caller to clean up */ - if (vfexec || (spawn && error == 0)) { - task_deallocate(new_task); - thread_deallocate(thread); + } else { + os_reason_free(exec_failure_reason); + exec_failure_reason = OS_REASON_NULL; } +done: if (load_result.threadstate) { kfree(load_result.threadstate, load_result.threadstate_sz); load_result.threadstate = NULL; } bad: - return(error); + /* If we hit this, we likely would have leaked an exit reason */ + assert(exec_failure_reason == OS_REASON_NULL); + return error; } @@ -1251,9 +1385,9 @@ struct execsw { int (*ex_imgact)(struct image_params *); const char *ex_name; } execsw[] = { - { exec_mach_imgact, "Mach-o Binary" }, - { exec_fat_imgact, "Fat Binary" }, - { exec_shell_imgact, "Interpreter Script" }, + { exec_mach_imgact, "Mach-o Binary" }, + { exec_fat_imgact, "Fat Binary" }, + { exec_shell_imgact, "Interpreter Script" }, { NULL, NULL} }; @@ -1263,7 +1397,8 @@ struct execsw { * * Description: Iterate through the available image activators, and activate * the image associated with the imgp structure. We start with - * the + * the activator for Mach-o binaries followed by that for Fat binaries + * for Interpreter scripts. * * Parameters: struct image_params * Image parameter block * @@ -1291,15 +1426,16 @@ exec_activate_image(struct image_params *imgp) const char *excpath; int error; int resid; - int once = 1; /* save SGUID-ness for interpreted files */ + int once = 1; /* save SGUID-ness for interpreted files */ int i; int itercount = 0; proc_t p = vfs_context_proc(imgp->ip_vfs_context); error = execargs_alloc(imgp); - if (error) + if (error) { goto bad_notrans; - + } + error = exec_save_path(imgp, imgp->ip_user_fname, imgp->ip_seg, &excpath); if (error) { goto bad_notrans; @@ -1315,14 +1451,15 @@ exec_activate_image(struct image_params *imgp) } NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1, - UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); + UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); again: error = namei(ndp); - if (error) + if (error) { goto bad_notrans; - imgp->ip_ndp = ndp; /* successful namei(); call nameidone() later */ - imgp->ip_vp = ndp->ni_vp; /* if set, need to vnode_put() at some point */ + } + imgp->ip_ndp = ndp; /* successful namei(); call nameidone() later */ + imgp->ip_vp = ndp->ni_vp; /* if set, need to vnode_put() at some point */ /* * Before we start the transition from binary A to binary B, make @@ -1339,12 +1476,14 @@ again: } error = proc_transstart(p, 1, 0); proc_unlock(p); - if (error) + if (error) { goto bad_notrans; + } error = exec_check_permissions(imgp); - if (error) + if (error) { goto bad; + } /* Copy; avoid invocation of an interpreter overwriting the original */ if (once) { @@ -1353,11 +1492,12 @@ again: } error = vn_rdwr(UIO_READ, imgp->ip_vp, imgp->ip_vdata, PAGE_SIZE, 0, - UIO_SYSSPACE, IO_NODELOCKED, - vfs_context_ucred(imgp->ip_vfs_context), - &resid, vfs_context_proc(imgp->ip_vfs_context)); - if (error) + UIO_SYSSPACE, IO_NODELOCKED, + vfs_context_ucred(imgp->ip_vfs_context), + &resid, vfs_context_proc(imgp->ip_vfs_context)); + if (error) { goto bad; + } if (resid) { memset(imgp->ip_vdata + (PAGE_SIZE - resid), 0x0, resid); @@ -1370,50 +1510,52 @@ encapsulated_binary: goto bad; } error = -1; - for(i = 0; error == -1 && execsw[i].ex_imgact != NULL; i++) { - + for (i = 0; error == -1 && execsw[i].ex_imgact != NULL; i++) { error = (*execsw[i].ex_imgact)(imgp); switch (error) { /* case -1: not claimed: continue */ - case -2: /* Encapsulated binary, imgp->ip_XXX set for next iteration */ + case -2: /* Encapsulated binary, imgp->ip_XXX set for next iteration */ goto encapsulated_binary; - case -3: /* Interpreter */ + case -3: /* Interpreter */ #if CONFIG_MACF /* * Copy the script label for later use. Note that * the label can be different when the script is * actually read by the interpreter. */ - if (imgp->ip_scriptlabelp) + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } imgp->ip_scriptlabelp = mac_vnode_label_alloc(); if (imgp->ip_scriptlabelp == NULL) { error = ENOMEM; break; } mac_vnode_label_copy(imgp->ip_vp->v_label, - imgp->ip_scriptlabelp); + imgp->ip_scriptlabelp); /* * Take a ref of the script vnode for later use. */ - if (imgp->ip_scriptvp) + if (imgp->ip_scriptvp) { vnode_put(imgp->ip_scriptvp); - if (vnode_getwithref(imgp->ip_vp) == 0) + } + if (vnode_getwithref(imgp->ip_vp) == 0) { imgp->ip_scriptvp = imgp->ip_vp; + } #endif nameidone(ndp); vnode_put(imgp->ip_vp); - imgp->ip_vp = NULL; /* already put */ + imgp->ip_vp = NULL; /* already put */ imgp->ip_ndp = NULL; /* already nameidone */ /* Use excpath, which exec_shell_imgact reset to the interpreter */ NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF, - UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); + UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); proc_transend(p, 0); goto again; @@ -1423,39 +1565,36 @@ encapsulated_binary: } } - /* - * Call out to allow 3rd party notification of exec. - * Ignore result of kauth_authorize_fileop call. - */ - if (error == 0 && kauth_authorize_fileop_has_listeners()) { - kauth_authorize_fileop(vfs_context_ucred(imgp->ip_vfs_context), - KAUTH_FILEOP_EXEC, - (uintptr_t)ndp->ni_vp, 0); - } - if (error == 0) { - /* - * Reset atm context from task - */ - task_atm_reset(p->task); + if (imgp->ip_flags & IMGPF_INTERPRET && ndp->ni_vp) { + AUDIT_ARG(vnpath, ndp->ni_vp, ARG_VNODE2); + } /* - * Reset old bank context from task + * Call out to allow 3rd party notification of exec. + * Ignore result of kauth_authorize_fileop call. */ - task_bank_reset(p->task); + if (kauth_authorize_fileop_has_listeners()) { + kauth_authorize_fileop(vfs_context_ucred(imgp->ip_vfs_context), + KAUTH_FILEOP_EXEC, + (uintptr_t)ndp->ni_vp, 0); + } } bad: proc_transend(p, 0); bad_notrans: - if (imgp->ip_strings) + if (imgp->ip_strings) { execargs_free(imgp); - if (imgp->ip_ndp) + } + if (imgp->ip_ndp) { nameidone(imgp->ip_ndp); - if (ndp) + } + if (ndp) { FREE(ndp, M_TEMP); + } - return (error); + return error; } @@ -1471,7 +1610,7 @@ bad_notrans: */ static errno_t exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, uint64_t psa_darwin_role, - ipc_port_t * portwatch_ports, int portwatch_count) + ipc_port_t * portwatch_ports, int portwatch_count) { int apptype = TASK_APPTYPE_NONE; int qos_clamp = THREAD_QOS_UNSPECIFIED; @@ -1480,47 +1619,49 @@ exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, if ((psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK) != 0) { int proctype = psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK; - switch(proctype) { - case POSIX_SPAWN_PROC_TYPE_DAEMON_INTERACTIVE: - apptype = TASK_APPTYPE_DAEMON_INTERACTIVE; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_STANDARD: - apptype = TASK_APPTYPE_DAEMON_STANDARD; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE: - apptype = TASK_APPTYPE_DAEMON_ADAPTIVE; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_BACKGROUND: - apptype = TASK_APPTYPE_DAEMON_BACKGROUND; - break; - case POSIX_SPAWN_PROC_TYPE_APP_DEFAULT: - apptype = TASK_APPTYPE_APP_DEFAULT; - break; - case POSIX_SPAWN_PROC_TYPE_APP_TAL: - apptype = TASK_APPTYPE_APP_TAL; - break; - default: - apptype = TASK_APPTYPE_NONE; - /* TODO: Should an invalid value here fail the spawn? */ - break; + switch (proctype) { + case POSIX_SPAWN_PROC_TYPE_DAEMON_INTERACTIVE: + apptype = TASK_APPTYPE_DAEMON_INTERACTIVE; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_STANDARD: + apptype = TASK_APPTYPE_DAEMON_STANDARD; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE: + apptype = TASK_APPTYPE_DAEMON_ADAPTIVE; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_BACKGROUND: + apptype = TASK_APPTYPE_DAEMON_BACKGROUND; + break; + case POSIX_SPAWN_PROC_TYPE_APP_DEFAULT: + apptype = TASK_APPTYPE_APP_DEFAULT; + break; +#if !CONFIG_EMBEDDED + case POSIX_SPAWN_PROC_TYPE_APP_TAL: + apptype = TASK_APPTYPE_APP_TAL; + break; +#endif /* !CONFIG_EMBEDDED */ + default: + apptype = TASK_APPTYPE_NONE; + /* TODO: Should an invalid value here fail the spawn? */ + break; } } if (psa_qos_clamp != POSIX_SPAWN_PROC_CLAMP_NONE) { switch (psa_qos_clamp) { - case POSIX_SPAWN_PROC_CLAMP_UTILITY: - qos_clamp = THREAD_QOS_UTILITY; - break; - case POSIX_SPAWN_PROC_CLAMP_BACKGROUND: - qos_clamp = THREAD_QOS_BACKGROUND; - break; - case POSIX_SPAWN_PROC_CLAMP_MAINTENANCE: - qos_clamp = THREAD_QOS_MAINTENANCE; - break; - default: - qos_clamp = THREAD_QOS_UNSPECIFIED; - /* TODO: Should an invalid value here fail the spawn? */ - break; + case POSIX_SPAWN_PROC_CLAMP_UTILITY: + qos_clamp = THREAD_QOS_UTILITY; + break; + case POSIX_SPAWN_PROC_CLAMP_BACKGROUND: + qos_clamp = THREAD_QOS_BACKGROUND; + break; + case POSIX_SPAWN_PROC_CLAMP_MAINTENANCE: + qos_clamp = THREAD_QOS_MAINTENANCE; + break; + default: + qos_clamp = THREAD_QOS_UNSPECIFIED; + /* TODO: Should an invalid value here fail the spawn? */ + break; } } @@ -1528,104 +1669,117 @@ exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, proc_darwin_role_to_task_role(psa_darwin_role, &role); } - if (apptype != TASK_APPTYPE_NONE || + if (apptype != TASK_APPTYPE_NONE || qos_clamp != THREAD_QOS_UNSPECIFIED || - role != TASK_UNSPECIFIED) { + role != TASK_UNSPECIFIED) { proc_set_task_spawnpolicy(p->task, apptype, qos_clamp, role, - portwatch_ports, portwatch_count); + portwatch_ports, portwatch_count); } - return (0); + return 0; } /* * exec_handle_port_actions * - * Description: Go through the _posix_port_actions_t contents, - * calling task_set_special_port, task_set_exception_ports - * and/or audit_session_spawnjoin for the current task. + * Description: Go through the _posix_port_actions_t contents, + * calling task_set_special_port, task_set_exception_ports + * and/or audit_session_spawnjoin for the current task. * * Parameters: struct image_params * Image parameter block - * short psa_flags posix spawn attribute flags * * Returns: 0 Success - * EINVAL Failure - * ENOTSUP Illegal posix_spawn attr flag was set + * EINVAL Failure + * ENOTSUP Illegal posix_spawn attr flag was set */ static errno_t -exec_handle_port_actions(struct image_params *imgp, short psa_flags, boolean_t * portwatch_present, ipc_port_t * portwatch_ports) +exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_present, + ipc_port_t * portwatch_ports) { _posix_spawn_port_actions_t pacts = imgp->ip_px_spa; +#if CONFIG_AUDIT proc_t p = vfs_context_proc(imgp->ip_vfs_context); +#endif _ps_port_action_t *act = NULL; - task_t task = p->task; + task_t task = get_threadtask(imgp->ip_new_thread); ipc_port_t port = NULL; errno_t ret = 0; int i; + kern_return_t kr; *portwatch_present = FALSE; for (i = 0; i < pacts->pspa_count; i++) { act = &pacts->pspa_actions[i]; - if (ipc_object_copyin(get_task_ipcspace(current_task()), - act->new_port, MACH_MSG_TYPE_COPY_SEND, - (ipc_object_t *) &port) != KERN_SUCCESS) { - ret = EINVAL; - goto done; + if (MACH_PORT_VALID(act->new_port)) { + kr = ipc_object_copyin(get_task_ipcspace(current_task()), + act->new_port, MACH_MSG_TYPE_COPY_SEND, + (ipc_object_t *) &port); + + if (kr != KERN_SUCCESS) { + ret = EINVAL; + goto done; + } + } else { + /* it's NULL or DEAD */ + port = CAST_MACH_NAME_TO_PORT(act->new_port); } switch (act->port_type) { case PSPA_SPECIAL: - /* Only allowed when not under vfork */ - if (!(psa_flags & POSIX_SPAWN_SETEXEC)) - ret = ENOTSUP; - else if (task_set_special_port(task, - act->which, port) != KERN_SUCCESS) + kr = task_set_special_port(task, act->which, port); + + if (kr != KERN_SUCCESS) { ret = EINVAL; + } break; case PSPA_EXCEPTION: - /* Only allowed when not under vfork */ - if (!(psa_flags & POSIX_SPAWN_SETEXEC)) - ret = ENOTSUP; - else if (task_set_exception_ports(task, - act->mask, port, act->behavior, - act->flavor) != KERN_SUCCESS) + kr = task_set_exception_ports(task, act->mask, port, + act->behavior, act->flavor); + if (kr != KERN_SUCCESS) { ret = EINVAL; + } break; #if CONFIG_AUDIT case PSPA_AU_SESSION: - ret = audit_session_spawnjoin(p, port); + ret = audit_session_spawnjoin(p, task, port); + if (ret) { + /* audit_session_spawnjoin() has already dropped the reference in case of error. */ + goto done; + } + break; #endif case PSPA_IMP_WATCHPORTS: - if (portwatch_ports != NULL) { + if (portwatch_ports != NULL && IPC_PORT_VALID(port)) { *portwatch_present = TRUE; /* hold on to this till end of spawn */ portwatch_ports[i] = port; - ret = 0; - } else + } else { ipc_port_release_send(port); + } + break; default: ret = EINVAL; break; } - /* action failed, so release port resources */ - - if (ret) { + if (ret) { + /* action failed, so release port resources */ ipc_port_release_send(port); break; } } done: - if (0 != ret) + if (0 != ret) { DTRACE_PROC1(spawn__port__failure, mach_port_name_t, act->new_port); - return (ret); + } + return ret; } /* @@ -1653,12 +1807,12 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) int action; proc_t p = vfs_context_proc(imgp->ip_vfs_context); _posix_spawn_file_actions_t px_sfap = imgp->ip_px_sfa; - int ival[2]; /* dummy retval for system calls) */ + int ival[2]; /* dummy retval for system calls) */ for (action = 0; action < px_sfap->psfa_act_count; action++) { - _psfa_action_t *psfa = &px_sfap->psfa_act_acts[ action]; + _psfa_action_t *psfa = &px_sfap->psfa_act_acts[action]; - switch(psfa->psfaa_type) { + switch (psfa->psfaa_type) { case PSFA_OPEN: { /* * Open is different, in that it requires the use of @@ -1687,19 +1841,19 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) VATTR_INIT(vap); /* Mask off all but regular access permissions */ - mode = ((mode &~ p->p_fd->fd_cmask) & ALLPERMS) & ~S_ISTXT; + mode = ((mode & ~p->p_fd->fd_cmask) & ALLPERMS) & ~S_ISTXT; VATTR_SET(vap, va_mode, mode & ACCESSPERMS); NDINIT(ndp, LOOKUP, OP_OPEN, FOLLOW | AUDITVNPATH1, UIO_SYSSPACE, - CAST_USER_ADDR_T(psfa->psfaa_openargs.psfao_path), - imgp->ip_vfs_context); + CAST_USER_ADDR_T(psfa->psfaa_openargs.psfao_path), + imgp->ip_vfs_context); - error = open1(imgp->ip_vfs_context, - ndp, - psfa->psfaa_openargs.psfao_oflag, - vap, - fileproc_alloc_init, NULL, - ival); + error = open1(imgp->ip_vfs_context, + ndp, + psfa->psfaa_openargs.psfao_oflag, + vap, + fileproc_alloc_init, NULL, + ival); FREE(bufp, M_TEMP); @@ -1709,8 +1863,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) * reworking all the open code to preallocate fd * slots, and internally taking one as an argument. */ - if (error || ival[0] == psfa->psfaa_filedes) + if (error || ival[0] == psfa->psfaa_filedes) { break; + } origfd = ival[0]; /* @@ -1728,8 +1883,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) * fd we wanted, the error will stop us. */ error = dup2(p, &dup2a, ival); - if (error) + if (error) { break; + } /* * Finally, close the original fd. @@ -1737,8 +1893,8 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) ca.fd = origfd; error = close_nocancel(p, &ca, ival); - } - break; + } + break; case PSFA_DUP2: { struct dup2_args dup2a; @@ -1753,8 +1909,8 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) * fd we wanted, the error will stop us. */ error = dup2(p, &dup2a, ival); - } - break; + } + break; case PSFA_CLOSE: { struct close_nocancel_args ca; @@ -1762,8 +1918,8 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) ca.fd = psfa->psfaa_filedes; error = close_nocancel(p, &ca, ival); - } - break; + } + break; case PSFA_INHERIT: { struct fcntl_nocancel_args fcntla; @@ -1777,8 +1933,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) */ fcntla.fd = psfa->psfaa_filedes; fcntla.cmd = F_GETFD; - if ((error = fcntl_nocancel(p, &fcntla, ival)) != 0) + if ((error = fcntl_nocancel(p, &fcntla, ival)) != 0) { break; + } if ((ival[0] & FD_CLOEXEC) == FD_CLOEXEC) { fcntla.fd = psfa->psfaa_filedes; @@ -1786,9 +1943,8 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) fcntla.arg = ival[0] & ~FD_CLOEXEC; error = fcntl_nocancel(p, &fcntla, ival); } - - } - break; + } + break; default: error = EINVAL; @@ -1800,7 +1956,7 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) if (error) { if (PSFA_OPEN == psfa->psfaa_type) { DTRACE_PROC1(spawn__open__failure, uintptr_t, - psfa->psfaa_openargs.psfao_path); + psfa->psfaa_openargs.psfao_path); } else { DTRACE_PROC1(spawn__fd__failure, int, psfa->psfaa_filedes); } @@ -1808,8 +1964,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) } } - if (error != 0 || (psa_flags & POSIX_SPAWN_CLOEXEC_DEFAULT) == 0) - return (error); + if (error != 0 || (psa_flags & POSIX_SPAWN_CLOEXEC_DEFAULT) == 0) { + return error; + } /* * If POSIX_SPAWN_CLOEXEC_DEFAULT is set, behave (during @@ -1828,7 +1985,7 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) switch (psfa->psfaa_type) { case PSFA_DUP2: fd = psfa->psfaa_openargs.psfao_oflag; - /*FALLTHROUGH*/ + /*FALLTHROUGH*/ case PSFA_OPEN: case PSFA_INHERIT: *fdflags(p, fd) |= UF_INHERIT; @@ -1840,7 +1997,7 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) } proc_fdunlock(p); - return (0); + return 0; } #if CONFIG_MACF @@ -1853,20 +2010,23 @@ exec_spawnattr_getmacpolicyinfo(const void *macextensions, const char *policynam const struct _posix_spawn_mac_policy_extensions *psmx = macextensions; int i; - if (psmx == NULL) + if (psmx == NULL) { return NULL; + } for (i = 0; i < psmx->psmx_count; i++) { const _ps_mac_policy_extension_t *extension = &psmx->psmx_extensions[i]; if (strncmp(extension->policyname, policyname, sizeof(extension->policyname)) == 0) { - if (lenp != NULL) + if (lenp != NULL) { *lenp = extension->datalen; + } return extension->datap; } } - if (lenp != NULL) + if (lenp != NULL) { *lenp = 0; + } return NULL; } @@ -1887,10 +2047,12 @@ spawn_copyin_macpolicyinfo(const struct user__posix_spawn_args_desc *px_args, _p } MALLOC(psmx, _posix_spawn_mac_policy_extensions_t, px_args->mac_extensions_size, M_TEMP, M_WAITOK); - if ((error = copyin(px_args->mac_extensions, psmx, px_args->mac_extensions_size)) != 0) + if ((error = copyin(px_args->mac_extensions, psmx, px_args->mac_extensions_size)) != 0) { goto bad; + } - if (PS_MAC_EXTENSIONS_SIZE(psmx->psmx_count) > px_args->mac_extensions_size) { + size_t extsize = PS_MAC_EXTENSIONS_SIZE(psmx->psmx_count); + if (extsize == 0 || extsize > px_args->mac_extensions_size) { error = EINVAL; goto bad; } @@ -1920,8 +2082,9 @@ spawn_copyin_macpolicyinfo(const struct user__posix_spawn_args_desc *px_args, _p bad: if (psmx != NULL) { - for (i = 0; i < copycnt; i++) + for (i = 0; i < copycnt; i++) { FREE(psmx->psmx_extensions[i].datap, M_TEMP); + } FREE(psmx, M_TEMP); } return error; @@ -1932,16 +2095,19 @@ spawn_free_macpolicyinfo(_posix_spawn_mac_policy_extensions_t psmx) { int i; - if (psmx == NULL) + if (psmx == NULL) { return; - for (i = 0; i < psmx->psmx_count; i++) + } + for (i = 0; i < psmx->psmx_count; i++) { FREE(psmx->psmx_extensions[i].datap, M_TEMP); + } FREE(psmx, M_TEMP); } #endif /* CONFIG_MACF */ #if CONFIG_COALITIONS -static inline void spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_TYPES]) +static inline void +spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_TYPES]) { for (int c = 0; c < COALITION_NUM_TYPES; c++) { if (coal[c]) { @@ -1953,7 +2119,8 @@ static inline void spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_T #endif #if CONFIG_PERSONAS -static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) +static int +spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) { int error = 0; struct persona *persona = NULL; @@ -1963,8 +2130,9 @@ static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) * TODO: rdar://problem/19981151 * Add entitlement check! */ - if (!kauth_cred_issuser(kauth_cred_get())) + if (!kauth_cred_issuser(kauth_cred_get())) { return EPERM; + } persona = persona_lookup(px_persona->pspi_id); if (!persona) { @@ -1986,15 +2154,15 @@ static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) } } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GROUPS) { - int ngroups = 0; + unsigned ngroups = 0; gid_t groups[NGROUPS_MAX]; if (persona_get_groups(persona, &ngroups, groups, - px_persona->pspi_ngroups) != 0) { + px_persona->pspi_ngroups) != 0) { error = EINVAL; goto out; } - if (ngroups != (int)px_persona->pspi_ngroups) { + if (ngroups != px_persona->pspi_ngroups) { error = EINVAL; goto out; } @@ -2012,29 +2180,33 @@ static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) } out: - if (persona) + if (persona) { persona_put(persona); + } return error; } -static int spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_persona) +static int +spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_persona) { int ret; kauth_cred_t cred; struct persona *persona = NULL; int override = !!(px_persona->pspi_flags & POSIX_SPAWN_PERSONA_FLAGS_OVERRIDE); - if (!override) + if (!override) { return persona_proc_adopt_id(p, px_persona->pspi_id, NULL); + } /* * we want to spawn into the given persona, but we want to override * the kauth with a different UID/GID combo */ persona = persona_lookup(px_persona->pspi_id); - if (!persona) + if (!persona) { return ESRCH; + } cred = persona_get_cred(persona); if (!cred) { @@ -2044,24 +2216,24 @@ static int spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_pe if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_UID) { cred = kauth_cred_setresuid(cred, - px_persona->pspi_uid, - px_persona->pspi_uid, - px_persona->pspi_uid, - KAUTH_UID_NONE); + px_persona->pspi_uid, + px_persona->pspi_uid, + px_persona->pspi_uid, + KAUTH_UID_NONE); } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GID) { cred = kauth_cred_setresgid(cred, - px_persona->pspi_gid, - px_persona->pspi_gid, - px_persona->pspi_gid); + px_persona->pspi_gid, + px_persona->pspi_gid, + px_persona->pspi_gid); } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GROUPS) { cred = kauth_cred_setgroups(cred, - px_persona->pspi_groups, - px_persona->pspi_ngroups, - px_persona->pspi_gmuid); + px_persona->pspi_groups, + px_persona->pspi_ngroups, + px_persona->pspi_gmuid); } ret = persona_proc_adopt(p, persona, cred); @@ -2072,49 +2244,21 @@ out: } #endif -void -proc_set_return_wait(proc_t p) -{ - proc_lock(p); - p->p_lflag |= P_LRETURNWAIT; - proc_unlock(p); -} - -void -proc_clear_return_wait(proc_t p, thread_t child_thread) -{ - proc_lock(p); - - p->p_lflag &= ~P_LRETURNWAIT; - if (p->p_lflag & P_LRETURNWAITER) { - wakeup(&p->p_lflag); - } - - proc_unlock(p); - - (void)thread_resume(child_thread); -} - -void -proc_wait_to_return() +#if __arm64__ +static inline void +proc_legacy_footprint(proc_t p, task_t task, const char *caller) { - proc_t p; - - p = current_proc(); - proc_lock(p); + boolean_t legacy_footprint_entitled; - if (p->p_lflag & P_LRETURNWAIT) { - p->p_lflag |= P_LRETURNWAITER; - do { - msleep(&p->p_lflag, &p->p_mlock, 0, - "thread_check_setup_complete", NULL); - } while (p->p_lflag & P_LRETURNWAIT); - p->p_lflag &= ~P_LRETURNWAITER; + legacy_footprint_entitled = IOTaskHasEntitlement(task, + "com.apple.private.memory.legacy_footprint"); + if (legacy_footprint_entitled) { + printf("%s: %d[%s] legacy footprint (entitled)\n", + caller, p->p_pid, p->p_name); + task_set_legacy_footprint(task, TRUE); } - - proc_unlock(p); - thread_bootstrap_return(); } +#endif /* __arm64__ */ /* * posix_spawn @@ -2147,14 +2291,14 @@ proc_wait_to_return() int posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) { - proc_t p = ap; /* quiet bogus GCC vfork() warning */ + proc_t p = ap; /* quiet bogus GCC vfork() warning */ user_addr_t pid = uap->pid; - int ival[2]; /* dummy retval for setpgid() */ - char *bufp = NULL; + int ival[2]; /* dummy retval for setpgid() */ + char *bufp = NULL; struct image_params *imgp; struct vnode_attr *vap; struct vnode_attr *origvap; - struct uthread *uthread = 0; /* compiler complains if not set to 0*/ + struct uthread *uthread = 0; /* compiler complains if not set to 0*/ int error, sig; int is_64 = IS_64BIT_PROCESS(p); struct vfs_context context; @@ -2168,13 +2312,17 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) boolean_t exec_done = FALSE; int portwatch_count = 0; ipc_port_t * portwatch_ports = NULL; - vm_size_t px_sa_offset = offsetof(struct _posix_spawnattr, psa_ports); + vm_size_t px_sa_offset = offsetof(struct _posix_spawnattr, psa_ports); + task_t old_task = current_task(); + task_t new_task = NULL; + boolean_t should_release_proc_ref = FALSE; + void *inherit = NULL; #if CONFIG_PERSONAS struct _posix_spawn_persona_info *px_persona = NULL; #endif /* - * Allocate a big chunk for locals instead of using stack since these + * Allocate a big chunk for locals instead of using stack since these * structures are pretty big. */ MALLOC(bufp, char *, (sizeof(*imgp) + sizeof(*vap) + sizeof(*origvap)), M_TEMP, M_WAITOK | M_ZERO); @@ -2193,13 +2341,14 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) imgp->ip_vattr = vap; imgp->ip_origvattr = origvap; imgp->ip_vfs_context = &context; - imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT : IMGPF_NONE); + imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT_ADDR : IMGPF_NONE); imgp->ip_seg = (is_64 ? UIO_USERSPACE64 : UIO_USERSPACE32); imgp->ip_mac_return = 0; imgp->ip_px_persona = NULL; + imgp->ip_cs_error = OS_REASON_NULL; if (uap->adesc != USER_ADDR_NULL) { - if(is_64) { + if (is_64) { error = copyin(uap->adesc, &px_args, sizeof(px_args)); } else { struct user32__posix_spawn_args_desc px_args32; @@ -2223,28 +2372,31 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) px_args.persona_info_size = px_args32.persona_info_size; px_args.persona_info = CAST_USER_ADDR_T(px_args32.persona_info); } - if (error) + if (error) { goto bad; + } if (px_args.attr_size != 0) { - /* - * We are not copying the port_actions pointer, - * because we already have it from px_args. + /* + * We are not copying the port_actions pointer, + * because we already have it from px_args. * This is a bit fragile: */ - if ((error = copyin(px_args.attrp, &px_sa, px_sa_offset) != 0)) - goto bad; - - bzero( (void *)( (unsigned long) &px_sa + px_sa_offset), sizeof(px_sa) - px_sa_offset ); + if ((error = copyin(px_args.attrp, &px_sa, px_sa_offset)) != 0) { + goto bad; + } + + bzero((void *)((unsigned long) &px_sa + px_sa_offset), sizeof(px_sa) - px_sa_offset ); imgp->ip_px_sa = &px_sa; } if (px_args.file_actions_size != 0) { /* Limit file_actions to allowed number of open files */ int maxfa = (p->p_limit ? p->p_rlimit[RLIMIT_NOFILE].rlim_cur : NOFILE); + size_t maxfa_size = PSF_ACTIONS_SIZE(maxfa); if (px_args.file_actions_size < PSF_ACTIONS_SIZE(1) || - px_args.file_actions_size > PSF_ACTIONS_SIZE(maxfa)) { + maxfa_size == 0 || px_args.file_actions_size > maxfa_size) { error = EINVAL; goto bad; } @@ -2255,12 +2407,14 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) } imgp->ip_px_sfa = px_sfap; - if ((error = copyin(px_args.file_actions, px_sfap, - px_args.file_actions_size)) != 0) + if ((error = copyin(px_args.file_actions, px_sfap, + px_args.file_actions_size)) != 0) { goto bad; + } /* Verify that the action count matches the struct size */ - if (PSF_ACTIONS_SIZE(px_sfap->psfa_act_count) != px_args.file_actions_size) { + size_t psfsize = PSF_ACTIONS_SIZE(px_sfap->psfa_act_count); + if (psfsize == 0 || psfsize != px_args.file_actions_size) { error = EINVAL; goto bad; } @@ -2268,25 +2422,27 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) if (px_args.port_actions_size != 0) { /* Limit port_actions to one page of data */ if (px_args.port_actions_size < PS_PORT_ACTIONS_SIZE(1) || - px_args.port_actions_size > PAGE_SIZE) { + px_args.port_actions_size > PAGE_SIZE) { error = EINVAL; goto bad; } - MALLOC(px_spap, _posix_spawn_port_actions_t, - px_args.port_actions_size, M_TEMP, M_WAITOK); + MALLOC(px_spap, _posix_spawn_port_actions_t, + px_args.port_actions_size, M_TEMP, M_WAITOK); if (px_spap == NULL) { error = ENOMEM; goto bad; } imgp->ip_px_spa = px_spap; - if ((error = copyin(px_args.port_actions, px_spap, - px_args.port_actions_size)) != 0) + if ((error = copyin(px_args.port_actions, px_spap, + px_args.port_actions_size)) != 0) { goto bad; + } /* Verify that the action count matches the struct size */ - if (PS_PORT_ACTIONS_SIZE(px_spap->pspa_count) != px_args.port_actions_size) { + size_t pasize = PS_PORT_ACTIONS_SIZE(px_spap->pspa_count); + if (pasize == 0 || pasize != px_args.port_actions_size) { error = EINVAL; goto bad; } @@ -2300,7 +2456,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) goto bad; } - MALLOC(px_persona, struct _posix_spawn_persona_info *, px_args.persona_info_size, M_TEMP, M_WAITOK|M_ZERO); + MALLOC(px_persona, struct _posix_spawn_persona_info *, px_args.persona_info_size, M_TEMP, M_WAITOK | M_ZERO); if (px_persona == NULL) { error = ENOMEM; goto bad; @@ -2308,16 +2464,19 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) imgp->ip_px_persona = px_persona; if ((error = copyin(px_args.persona_info, px_persona, - px_args.persona_info_size)) != 0) + px_args.persona_info_size)) != 0) { goto bad; - if ((error = spawn_validate_persona(px_persona)) != 0) + } + if ((error = spawn_validate_persona(px_persona)) != 0) { goto bad; + } } #endif #if CONFIG_MACF if (px_args.mac_extensions_size != 0) { - if ((error = spawn_copyin_macpolicyinfo(&px_args, (_posix_spawn_mac_policy_extensions_t *)&imgp->ip_px_smpx)) != 0) + if ((error = spawn_copyin_macpolicyinfo(&px_args, (_posix_spawn_mac_policy_extensions_t *)&imgp->ip_px_smpx)) != 0) { goto bad; + } } #endif /* CONFIG_MACF */ } @@ -2331,8 +2490,8 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) * which depends on it. */ if (uthread->uu_flag & UT_VFORK) { - error = EINVAL; - goto bad; + error = EINVAL; + goto bad; } /* @@ -2342,8 +2501,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) * which is one of the most expensive things about using fork() * and execve(). */ - if (imgp->ip_px_sa == NULL || !(px_sa.psa_flags & POSIX_SPAWN_SETEXEC)){ - + if (imgp->ip_px_sa == NULL || !(px_sa.psa_flags & POSIX_SPAWN_SETEXEC)) { /* Set the new task's coalition, if it is requested. */ coalition_t coal[COALITION_NUM_TYPES] = { COALITION_NULL }; #if CONFIG_COALITIONS @@ -2352,17 +2510,20 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) struct _posix_spawn_coalition_info coal_info; int coal_role[COALITION_NUM_TYPES]; - if (imgp->ip_px_sa == NULL || !px_args.coal_info) + if (imgp->ip_px_sa == NULL || !px_args.coal_info) { goto do_fork1; + } memset(&coal_info, 0, sizeof(coal_info)); - if (px_args.coal_info_size > sizeof(coal_info)) + if (px_args.coal_info_size > sizeof(coal_info)) { px_args.coal_info_size = sizeof(coal_info); + } error = copyin(px_args.coal_info, - &coal_info, px_args.coal_info_size); - if (error != 0) + &coal_info, px_args.coal_info_size); + if (error != 0) { goto bad; + } ncoals = 0; for (i = 0; i < COALITION_NUM_TYPES; i++) { @@ -2375,8 +2536,8 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) */ if (!task_is_in_privileged_coalition(p->task, i)) { coal_dbg("ERROR: %d not in privilegd " - "coalition of type %d", - p->p_pid, i); + "coalition of type %d", + p->p_pid, i); spawn_coalitions_release_all(coal); error = EPERM; goto bad; @@ -2391,7 +2552,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) coal[i] = coalition_find_and_activate_by_id(cid); if (coal[i] == COALITION_NULL) { coal_dbg("could not find coalition id:%llu " - "(perhaps it has been terminated or reaped)", cid); + "(perhaps it has been terminated or reaped)", cid); /* * release any other coalition's we * may have a reference to @@ -2402,7 +2563,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) } if (coalition_type(coal[i]) != i) { coal_dbg("coalition with id:%lld is not of type:%d" - " (it's type:%d)", cid, i, coalition_type(coal[i])); + " (it's type:%d)", cid, i, coalition_type(coal[i])); error = ESRCH; goto bad; } @@ -2434,13 +2595,34 @@ do_fork1: * caller's persona (if it exists) */ error = fork1(p, &imgp->ip_new_thread, PROC_CREATE_SPAWN, coal); + /* returns a thread and task reference */ + if (error == 0) { + new_task = get_threadtask(imgp->ip_new_thread); + } #if CONFIG_COALITIONS /* set the roles of this task within each given coalition */ if (error == 0) { - kr = coalitions_set_roles(coal, get_threadtask(imgp->ip_new_thread), coal_role); - if (kr != KERN_SUCCESS) + kr = coalitions_set_roles(coal, new_task, coal_role); + if (kr != KERN_SUCCESS) { error = EINVAL; + } + if (kdebug_debugid_enabled(MACHDBG_CODE(DBG_MACH_COALITION, + MACH_COALITION_ADOPT))) { + for (i = 0; i < COALITION_NUM_TYPES; i++) { + if (coal[i] != COALITION_NULL) { + /* + * On 32-bit targets, uniqueid + * will get truncated to 32 bits + */ + KDBG_RELEASE(MACHDBG_CODE( + DBG_MACH_COALITION, + MACH_COALITION_ADOPT), + coalition_id(coal[i]), + get_task_uniqueid(new_task)); + } + } + } } /* drop our references and activations - fork1() now holds them */ @@ -2449,8 +2631,8 @@ do_fork1: if (error != 0) { goto bad; } - imgp->ip_flags |= IMGPF_SPAWN; /* spawn w/o exec */ - spawn_no_exec = TRUE; /* used in later tests */ + imgp->ip_flags |= IMGPF_SPAWN; /* spawn w/o exec */ + spawn_no_exec = TRUE; /* used in later tests */ #if CONFIG_PERSONAS /* @@ -2466,7 +2648,7 @@ do_fork1: #if 0 if (!proc_has_persona(p) && imgp->ip_px_persona == NULL) { MALLOC(px_persona, struct _posix_spawn_persona_info *, - sizeof(*px_persona), M_TEMP, M_WAITOK|M_ZERO); + sizeof(*px_persona), M_TEMP, M_WAITOK | M_ZERO); if (px_persona == NULL) { error = ENOMEM; goto bad; @@ -2476,11 +2658,52 @@ do_fork1: } #endif /* 0 */ #endif /* CONFIG_PERSONAS */ + } else { + /* + * For execve case, create a new task and thread + * which points to current_proc. The current_proc will point + * to the new task after image activation and proc ref drain. + * + * proc (current_proc) <----- old_task (current_task) + * ^ | ^ + * | | | + * | ---------------------------------- + * | + * --------- new_task (task marked as TF_EXEC_COPY) + * + * After image activation, the proc will point to the new task + * and would look like following. + * + * proc (current_proc) <----- old_task (current_task, marked as TPF_DID_EXEC) + * ^ | + * | | + * | ----------> new_task + * | | + * ----------------- + * + * During exec any transition from new_task -> proc is fine, but don't allow + * transition from proc->task, since it will modify old_task. + */ + imgp->ip_new_thread = fork_create_child(old_task, + NULL, + p, + FALSE, + p->p_flag & P_LP64, + task_get_64bit_data(old_task), + TRUE); + /* task and thread ref returned by fork_create_child */ + if (imgp->ip_new_thread == NULL) { + error = ENOMEM; + goto bad; + } + + new_task = get_threadtask(imgp->ip_new_thread); + imgp->ip_flags |= IMGPF_EXEC; } if (spawn_no_exec) { p = (proc_t)get_bsdthreadtask_info(imgp->ip_new_thread); - + /* * We had to wait until this point before firing the * proc:::create probe, otherwise p would not point to the @@ -2490,16 +2713,8 @@ do_fork1: } assert(p != NULL); - /* By default, the thread everyone plays with is the parent */ - context.vc_thread = current_thread(); - context.vc_ucred = p->p_ucred; /* XXX must NOT be kauth_cred_get() */ - - /* - * However, if we're not in the setexec case, redirect the context - * to the newly created process instead - */ - if (spawn_no_exec) - context.vc_thread = imgp->ip_new_thread; + context.vc_thread = imgp->ip_new_thread; + context.vc_ucred = p->p_ucred; /* XXX must NOT be kauth_cred_get() */ /* * Post fdcopy(), pre exec_handle_sugid() - this is where we want @@ -2515,8 +2730,9 @@ do_fork1: * is handled in exec_handle_file_actions(). */ if ((error = exec_handle_file_actions(imgp, - imgp->ip_px_sa != NULL ? px_sa.psa_flags : 0)) != 0) + imgp->ip_px_sa != NULL ? px_sa.psa_flags : 0)) != 0) { goto bad; + } } /* Has spawn port actions? */ @@ -2525,8 +2741,9 @@ do_fork1: boolean_t portwatch_present = FALSE; /* Will this process become adaptive? The apptype isn't ready yet, so we can't look there. */ - if (imgp->ip_px_sa != NULL && px_sa.psa_apptype == POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE) + if (imgp->ip_px_sa != NULL && px_sa.psa_apptype == POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE) { is_adaptive = TRUE; + } /* * portwatch only: @@ -2540,9 +2757,9 @@ do_fork1: portwatch_ports = NULL; } - if ((error = exec_handle_port_actions(imgp, - imgp->ip_px_sa != NULL ? px_sa.psa_flags : 0, &portwatch_present, portwatch_ports)) != 0) + if ((error = exec_handle_port_actions(imgp, &portwatch_present, portwatch_ports)) != 0) { goto bad; + } if (portwatch_present == FALSE && portwatch_ports != NULL) { FREE(portwatch_ports, M_TEMP); @@ -2565,8 +2782,9 @@ do_fork1: * Effectively, call setpgid() system call; works * because there are no pointer arguments. */ - if((error = setpgid(p, &spga, ival)) != 0) + if ((error = setpgid(p, &spga, ival)) != 0) { goto bad; + } } /* @@ -2578,18 +2796,35 @@ do_fork1: * image will take precedence over the spawn attributes * (re)setting them. * - * The use of p_ucred is safe, since we are acting on the - * new process, and it has no threads other than the one - * we are creating for it. + * Modifications to p_ucred must be guarded using the + * proc's ucred lock. This prevents others from accessing + * a garbage credential. */ - if (px_sa.psa_flags & POSIX_SPAWN_RESETIDS) { - kauth_cred_t my_cred = p->p_ucred; + while (px_sa.psa_flags & POSIX_SPAWN_RESETIDS) { + kauth_cred_t my_cred = kauth_cred_proc_ref(p); kauth_cred_t my_new_cred = kauth_cred_setuidgid(my_cred, kauth_cred_getruid(my_cred), kauth_cred_getrgid(my_cred)); - if (my_new_cred != my_cred) { - p->p_ucred = my_new_cred; - /* update cred on proc */ - PROC_UPDATE_CREDS_ONPROC(p); + + if (my_cred == my_new_cred) { + kauth_cred_unref(&my_cred); + break; + } + + /* update cred on proc */ + proc_ucred_lock(p); + + if (p->p_ucred != my_cred) { + proc_ucred_unlock(p); + kauth_cred_unref(&my_new_cred); + continue; } + + /* donate cred reference on my_new_cred to p->p_ucred */ + p->p_ucred = my_new_cred; + PROC_UPDATE_CREDS_ONPROC(p); + proc_ucred_unlock(p); + + /* drop additional reference that was taken on the previous cred */ + kauth_cred_unref(&my_cred); } #if CONFIG_PERSONAS @@ -2603,8 +2838,9 @@ do_fork1: * being executed. */ error = spawn_persona_adopt(p, imgp->ip_px_persona); - if (error != 0) + if (error != 0) { goto bad; + } } #endif /* CONFIG_PERSONAS */ #if !SECURE_KERNEL @@ -2618,16 +2854,23 @@ do_fork1: * useful or necessary to disable ASLR on a per-process * basis for unit testing and debugging. */ - if (px_sa.psa_flags & _POSIX_SPAWN_DISABLE_ASLR) + if (px_sa.psa_flags & _POSIX_SPAWN_DISABLE_ASLR) { OSBitOrAtomic(P_DISABLE_ASLR, &p->p_flag); + } #endif /* !SECURE_KERNEL */ + /* Randomize high bits of ASLR slide */ + if (px_sa.psa_flags & _POSIX_SPAWN_HIGH_BITS_ASLR) { + imgp->ip_flags |= IMGPF_HIGH_BITS_ASLR; + } + /* * Forcibly disallow execution from data pages for the spawned process * even if it would otherwise be permitted by the architecture default. */ - if (px_sa.psa_flags & _POSIX_SPAWN_ALLOW_DATA_EXEC) + if (px_sa.psa_flags & _POSIX_SPAWN_ALLOW_DATA_EXEC) { imgp->ip_flags |= IMGPF_ALLOW_DATA_EXEC; + } } /* @@ -2635,10 +2878,11 @@ do_fork1: * _POSIX_SPAWN_DISABLE_ASLR attribute was found above or if * P_DISABLE_ASLR was inherited from the parent process. */ - if (p->p_flag & P_DISABLE_ASLR) + if (p->p_flag & P_DISABLE_ASLR) { imgp->ip_flags |= IMGPF_DISABLE_ASLR; + } - /* + /* * Clear transition flag so we won't hang if exec_activate_image() causes * an automount (and launchd does a proc sysctl to service it). * @@ -2649,11 +2893,12 @@ do_fork1: proc_transit_set = 0; } -#if MAC_SPAWN /* XXX */ +#if MAC_SPAWN /* XXX */ if (uap->mac_p != USER_ADDR_NULL) { error = mac_execve_enter(uap->mac_p, imgp); - if (error) + if (error) { goto bad; + } } #endif @@ -2661,7 +2906,22 @@ do_fork1: * Activate the image */ error = exec_activate_image(imgp); - + + if (error == 0 && !spawn_no_exec) { + p = proc_exec_switch_task(p, old_task, new_task, imgp->ip_new_thread); + /* proc ref returned */ + should_release_proc_ref = TRUE; + + /* + * Need to transfer pending watch port boosts to the new task while still making + * sure that the old task remains in the importance linkage. Create an importance + * linkage from old task to new task, then switch the task importance base + * of old task and new task. After the switch the port watch boost will be + * boosting the new task and new task will be donating importance to old task. + */ + inherit = ipc_importance_exec_switch_task(old_task, new_task); + } + if (error == 0) { /* process completed the exec */ exec_done = TRUE; @@ -2680,26 +2940,17 @@ do_fork1: * until after the image is activated. */ if (!error && imgp->ip_px_sa != NULL) { - thread_t child_thread = current_thread(); - uthread_t child_uthread = uthread; - - /* - * If we created a new child thread, then the thread and - * uthread are different than the current ones; otherwise, - * we leave them, since we are in the exec case instead. - */ - if (spawn_no_exec) { - child_thread = imgp->ip_new_thread; - child_uthread = get_bsdthread_info(child_thread); - } + thread_t child_thread = imgp->ip_new_thread; + uthread_t child_uthread = get_bsdthread_info(child_thread); /* * Mask a list of signals, instead of them being unmasked, if * they were unmasked in the parent; note that some signals * are not maskable. */ - if (px_sa.psa_flags & POSIX_SPAWN_SETSIGMASK) + if (px_sa.psa_flags & POSIX_SPAWN_SETSIGMASK) { child_uthread->uu_sigmask = (px_sa.psa_sigmask & ~sigcantmask); + } /* * Default a list of signals instead of ignoring them, if * they were ignored in the parent. Note that we pass @@ -2712,9 +2963,10 @@ do_fork1: vec.sa_tramp = 0; vec.sa_mask = 0; vec.sa_flags = 0; - for (sig = 0; sig < NSIG; sig++) - if (px_sa.psa_sigdefault & (1 << sig)) { - error = setsigvec(p, child_thread, sig + 1, &vec, spawn_no_exec); + for (sig = 1; sig < NSIG; sig++) { + if (px_sa.psa_sigdefault & (1 << (sig - 1))) { + error = setsigvec(p, child_thread, sig, &vec, spawn_no_exec); + } } } @@ -2732,10 +2984,10 @@ do_fork1: * whomever is turning it on could just as easily choose not to do so. */ error = proc_set_task_ruse_cpu(p->task, - TASK_POLICY_RESOURCE_ATTRIBUTE_NOTIFY_EXC, - px_sa.psa_cpumonitor_percent, - px_sa.psa_cpumonitor_interval * NSEC_PER_SEC, - 0, TRUE); + TASK_POLICY_RESOURCE_ATTRIBUTE_NOTIFY_EXC, + px_sa.psa_cpumonitor_percent, + px_sa.psa_cpumonitor_interval * NSEC_PER_SEC, + 0, TRUE); } } @@ -2743,61 +2995,63 @@ bad: if (error == 0) { /* reset delay idle sleep status if set */ - if ((p->p_flag & P_DELAYIDLESLEEP) == P_DELAYIDLESLEEP) +#if !CONFIG_EMBEDDED + if ((p->p_flag & P_DELAYIDLESLEEP) == P_DELAYIDLESLEEP) { OSBitAndAtomic(~((uint32_t)P_DELAYIDLESLEEP), &p->p_flag); + } +#endif /* !CONFIG_EMBEDDED */ /* upon successful spawn, re/set the proc control state */ if (imgp->ip_px_sa != NULL) { switch (px_sa.psa_pcontrol) { - case POSIX_SPAWN_PCONTROL_THROTTLE: - p->p_pcaction = P_PCTHROTTLE; - break; - case POSIX_SPAWN_PCONTROL_SUSPEND: - p->p_pcaction = P_PCSUSP; - break; - case POSIX_SPAWN_PCONTROL_KILL: - p->p_pcaction = P_PCKILL; - break; - case POSIX_SPAWN_PCONTROL_NONE: - default: - p->p_pcaction = 0; - break; - }; + case POSIX_SPAWN_PCONTROL_THROTTLE: + p->p_pcaction = P_PCTHROTTLE; + break; + case POSIX_SPAWN_PCONTROL_SUSPEND: + p->p_pcaction = P_PCSUSP; + break; + case POSIX_SPAWN_PCONTROL_KILL: + p->p_pcaction = P_PCKILL; + break; + case POSIX_SPAWN_PCONTROL_NONE: + default: + p->p_pcaction = 0; + break; + } + ; } exec_resettextvp(p, imgp); - -#if CONFIG_MEMORYSTATUS && CONFIG_JETSAM + +#if CONFIG_MEMORYSTATUS /* Has jetsam attributes? */ if (imgp->ip_px_sa != NULL && (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_SET)) { /* * With 2-level high-water-mark support, POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND is no * longer relevant, as background limits are described via the inactive limit slots. - * At the kernel layer, the flag is ignored. * * That said, however, if the POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND is passed in, * we attempt to mimic previous behavior by forcing the BG limit data into the * inactive/non-fatal mode and force the active slots to hold system_wide/fatal mode. - * The kernel layer will flag this mapping. */ if (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND) { memorystatus_update(p, px_sa.psa_priority, 0, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), - TRUE, - -1, TRUE, - px_sa.psa_memlimit_inactive, FALSE, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND)); + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), + TRUE, + -1, TRUE, + px_sa.psa_memlimit_inactive, FALSE); } else { memorystatus_update(p, px_sa.psa_priority, 0, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), - TRUE, - px_sa.psa_memlimit_active, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_ACTIVE_FATAL), - px_sa.psa_memlimit_inactive, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_INACTIVE_FATAL), - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND)); + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), + TRUE, + px_sa.psa_memlimit_active, + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_ACTIVE_FATAL), + px_sa.psa_memlimit_inactive, + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_INACTIVE_FATAL)); } - } -#endif /* CONFIG_MEMORYSTATUS && CONFIG_JETSAM*/ +#endif /* CONFIG_MEMORYSTATUS */ + if (imgp->ip_px_sa != NULL && px_sa.psa_thread_limit > 0) { + task_set_thread_limit(new_task, (uint16_t)px_sa.psa_thread_limit); + } } /* @@ -2809,8 +3063,9 @@ bad: * before check_for_signature(), which uses psignal. */ if (spawn_no_exec) { - if (proc_transit_set) + if (proc_transit_set) { proc_transend(p, 0); + } /* * Drop the signal lock on the child which was taken on our @@ -2821,15 +3076,14 @@ bad: /* flag the 'fork' has occurred */ proc_knote(p->p_pptr, NOTE_FORK | p->p_pid); - /* then flag exec has occurred */ - /* notify only if it has not failed due to FP Key error */ - if ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0) - proc_knote(p, NOTE_EXEC); - } else if (error == 0) { - /* reset the importance attribute from our previous life */ - task_importance_reset(p->task); } + /* flag exec has occurred, notify only if it has not failed due to FP Key error */ + if (!error && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) { + proc_knote(p, NOTE_EXEC); + } + + if (error == 0) { /* * We need to initialize the bank context behind the protection of @@ -2840,11 +3094,19 @@ bad: error = proc_transstart(p, 0, 0); if (error == 0) { - task_bank_init(p->task); + task_bank_init(new_task); proc_transend(p, 0); } + +#if __arm64__ + proc_legacy_footprint(p, new_task, __FUNCTION__); +#endif /* __arm64__ */ } + /* Inherit task role from old task to new task for exec */ + if (error == 0 && !spawn_no_exec) { + proc_inherit_task_role(new_task, old_task); + } /* * Apply the spawnattr policy, apptype (which primes the task for importance donation), @@ -2859,14 +3121,34 @@ bad: struct _posix_spawnattr *psa = (struct _posix_spawnattr *) imgp->ip_px_sa; exec_handle_spawnattr_policy(p, psa->psa_apptype, psa->psa_qos_clamp, psa->psa_darwin_role, - portwatch_ports, portwatch_count); + portwatch_ports, portwatch_count); + } + + /* + * Apply the requested maximum address. + */ + if (error == 0 && imgp->ip_px_sa != NULL) { + struct _posix_spawnattr *psa = (struct _posix_spawnattr *) imgp->ip_px_sa; + + if (psa->psa_max_addr) { + vm_map_set_max_addr(get_task_map(new_task), psa->psa_max_addr); + } } - /* Apply the main thread qos */ if (error == 0) { - thread_t main_thread = (imgp->ip_new_thread != NULL) ? imgp->ip_new_thread : current_thread(); + /* Apply the main thread qos */ + thread_t main_thread = imgp->ip_new_thread; + task_set_main_thread_qos(new_task, main_thread); - task_set_main_thread_qos(p->task, main_thread); +#if CONFIG_MACF + /* + * Processes with the MAP_JIT entitlement are permitted to have + * a jumbo-size map. + */ + if (mac_proc_check_map_anon(p, 0, 0, 0, MAP_JIT, NULL) == 0) { + vm_map_set_jumbo(get_task_map(new_task)); + } +#endif /* CONFIG_MACF */ } /* @@ -2901,31 +3183,49 @@ bad: if (imgp != NULL && spawn_no_exec && (p->p_lflag & P_LTRACED)) { psignal_vfork(p, p->task, imgp->ip_new_thread, SIGTRAP); } + + if (error == 0 && !spawn_no_exec) { + KDBG(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXEC), + p->p_pid); + } } if (imgp != NULL) { - if (imgp->ip_vp) + if (imgp->ip_vp) { vnode_put(imgp->ip_vp); - if (imgp->ip_scriptvp) + } + if (imgp->ip_scriptvp) { vnode_put(imgp->ip_scriptvp); - if (imgp->ip_strings) + } + if (imgp->ip_strings) { execargs_free(imgp); - if (imgp->ip_px_sfa != NULL) + } + if (imgp->ip_px_sfa != NULL) { FREE(imgp->ip_px_sfa, M_TEMP); - if (imgp->ip_px_spa != NULL) + } + if (imgp->ip_px_spa != NULL) { FREE(imgp->ip_px_spa, M_TEMP); + } #if CONFIG_PERSONAS - if (imgp->ip_px_persona != NULL) + if (imgp->ip_px_persona != NULL) { FREE(imgp->ip_px_persona, M_TEMP); + } #endif #if CONFIG_MACF - if (imgp->ip_px_smpx != NULL) + if (imgp->ip_px_smpx != NULL) { spawn_free_macpolicyinfo(imgp->ip_px_smpx); - if (imgp->ip_execlabelp) + } + if (imgp->ip_execlabelp) { mac_cred_label_free(imgp->ip_execlabelp); - if (imgp->ip_scriptlabelp) + } + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } + if (imgp->ip_cs_error != OS_REASON_NULL) { + os_reason_free(imgp->ip_cs_error); + imgp->ip_cs_error = OS_REASON_NULL; + } #endif } @@ -2963,21 +3263,46 @@ bad: if (error) { DTRACE_PROC1(exec__failure, int, error); } else { - DTRACE_PROC(exec__success); + dtrace_thread_didexec(imgp->ip_new_thread); } } - if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) + if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) { (*dtrace_proc_waitfor_hook)(p); + } +#endif + +#if CONFIG_AUDIT + if (!error && AUDIT_ENABLED() && p) { + /* Add the CDHash of the new process to the audit record */ + uint8_t *cdhash = cs_get_cdhash(p); + if (cdhash) { + AUDIT_ARG(data, cdhash, sizeof(uint8_t), CS_CDHASH_LEN); + } + } #endif + /* + * clear bsd_info from old task if it did exec. + */ + if (task_did_exec(old_task)) { + set_bsdtask_info(old_task, NULL); + } + + /* clear bsd_info from new task and terminate it if exec failed */ + if (new_task != NULL && task_is_exec_copy(new_task)) { + set_bsdtask_info(new_task, NULL); + task_terminate_internal(new_task); + } + /* Return to both the parent and the child? */ if (imgp != NULL && spawn_no_exec) { /* * If the parent wants the pid, copy it out */ - if (pid != USER_ADDR_NULL) + if (pid != USER_ADDR_NULL) { (void)suword(pid, p->p_pid); + } retval[0] = error; /* @@ -2994,37 +3319,177 @@ bad: p->exit_thread = current_thread(); proc_unlock(p); exit1(p, 1, (int *)NULL); - proc_clear_return_wait(p, imgp->ip_new_thread); - if (exec_done == FALSE) { - task_deallocate(get_threadtask(imgp->ip_new_thread)); - thread_deallocate(imgp->ip_new_thread); - } } else { /* someone is doing it for us; just skip it */ proc_unlock(p); - proc_clear_return_wait(p, imgp->ip_new_thread); } - } else { - - /* - * Return to the child - * - * Note: the image activator earlier dropped the - * task/thread references to the newly spawned - * process; this is OK, since we still have suspended - * queue references on them, so we should be fine - * with the delayed resume of the thread here. - */ - proc_clear_return_wait(p, imgp->ip_new_thread); } } + + /* + * Do not terminate the current task, if proc_exec_switch_task did not + * switch the tasks, terminating the current task without the switch would + * result in loosing the SIGKILL status. + */ + if (task_did_exec(old_task)) { + /* Terminate the current task, since exec will start in new task */ + task_terminate_internal(old_task); + } + + /* Release the thread ref returned by fork_create_child/fork1 */ + if (imgp != NULL && imgp->ip_new_thread) { + /* wake up the new thread */ + task_clear_return_wait(get_threadtask(imgp->ip_new_thread)); + thread_deallocate(imgp->ip_new_thread); + imgp->ip_new_thread = NULL; + } + + /* Release the ref returned by fork_create_child/fork1 */ + if (new_task) { + task_deallocate(new_task); + new_task = NULL; + } + + if (should_release_proc_ref) { + proc_rele(p); + } + if (bufp != NULL) { FREE(bufp, M_TEMP); } - - return(error); + + if (inherit != NULL) { + ipc_importance_release(inherit); + } + + return error; } +/* + * proc_exec_switch_task + * + * Parameters: p proc + * old_task task before exec + * new_task task after exec + * new_thread thread in new task + * + * Returns: proc. + * + * Note: The function will switch the task pointer of proc + * from old task to new task. The switch needs to happen + * after draining all proc refs and inside a proc translock. + * In the case of failure to switch the task, which might happen + * if the process received a SIGKILL or jetsam killed it, it will make + * sure that the new tasks terminates. User proc ref returned + * to caller. + * + * This function is called after point of no return, in the case + * failure to switch, it will terminate the new task and swallow the + * error and let the terminated process complete exec and die. + */ +proc_t +proc_exec_switch_task(proc_t p, task_t old_task, task_t new_task, thread_t new_thread) +{ + int error = 0; + boolean_t task_active; + boolean_t proc_active; + boolean_t thread_active; + thread_t old_thread = current_thread(); + + /* + * Switch the task pointer of proc to new task. + * Before switching the task, wait for proc_refdrain. + * After the switch happens, the proc can disappear, + * take a ref before it disappears. Waiting for + * proc_refdrain in exec will block all other threads + * trying to take a proc ref, boost the current thread + * to avoid priority inversion. + */ + thread_set_exec_promotion(old_thread); + p = proc_refdrain_with_refwait(p, TRUE); + /* extra proc ref returned to the caller */ + + assert(get_threadtask(new_thread) == new_task); + task_active = task_is_active(new_task); + + /* Take the proc_translock to change the task ptr */ + proc_lock(p); + proc_active = !(p->p_lflag & P_LEXIT); + + /* Check if the current thread is not aborted due to SIGKILL */ + thread_active = thread_is_active(old_thread); + + /* + * Do not switch the task if the new task or proc is already terminated + * as a result of error in exec past point of no return + */ + if (proc_active && task_active && thread_active) { + error = proc_transstart(p, 1, 0); + if (error == 0) { + uthread_t new_uthread = get_bsdthread_info(new_thread); + uthread_t old_uthread = get_bsdthread_info(current_thread()); + + /* + * bsd_info of old_task will get cleared in execve and posix_spawn + * after firing exec-success/error dtrace probe. + */ + p->task = new_task; + + /* Clear dispatchqueue and workloop ast offset */ + p->p_dispatchqueue_offset = 0; + p->p_dispatchqueue_serialno_offset = 0; + p->p_return_to_kernel_offset = 0; + + /* Copy the signal state, dtrace state and set bsd ast on new thread */ + act_set_astbsd(new_thread); + new_uthread->uu_siglist = old_uthread->uu_siglist; + new_uthread->uu_sigwait = old_uthread->uu_sigwait; + new_uthread->uu_sigmask = old_uthread->uu_sigmask; + new_uthread->uu_oldmask = old_uthread->uu_oldmask; + new_uthread->uu_vforkmask = old_uthread->uu_vforkmask; + new_uthread->uu_exit_reason = old_uthread->uu_exit_reason; +#if CONFIG_DTRACE + new_uthread->t_dtrace_sig = old_uthread->t_dtrace_sig; + new_uthread->t_dtrace_stop = old_uthread->t_dtrace_stop; + new_uthread->t_dtrace_resumepid = old_uthread->t_dtrace_resumepid; + assert(new_uthread->t_dtrace_scratch == NULL); + new_uthread->t_dtrace_scratch = old_uthread->t_dtrace_scratch; + + old_uthread->t_dtrace_sig = 0; + old_uthread->t_dtrace_stop = 0; + old_uthread->t_dtrace_resumepid = 0; + old_uthread->t_dtrace_scratch = NULL; +#endif + /* Copy the resource accounting info */ + thread_copy_resource_info(new_thread, current_thread()); + + /* Clear the exit reason and signal state on old thread */ + old_uthread->uu_exit_reason = NULL; + old_uthread->uu_siglist = 0; + + /* Add the new uthread to proc uthlist and remove the old one */ + TAILQ_INSERT_TAIL(&p->p_uthlist, new_uthread, uu_list); + TAILQ_REMOVE(&p->p_uthlist, old_uthread, uu_list); + + task_set_did_exec_flag(old_task); + task_clear_exec_copy_flag(new_task); + + task_copy_fields_for_exec(new_task, old_task); + + proc_transend(p, 1); + } + } + + proc_unlock(p); + proc_refwake(p); + thread_clear_exec_promotion(old_thread); + + if (error != 0 || !task_active || !proc_active || !thread_active) { + task_terminate_internal(new_task); + } + + return p; +} /* * execve @@ -3062,7 +3527,7 @@ execve(proc_t p, struct execve_args *uap, int32_t *retval) muap.mac_p = USER_ADDR_NULL; err = __mac_execve(p, &muap, retval); - return(err); + return err; } /* @@ -3094,19 +3559,25 @@ execve(proc_t p, struct execve_args *uap, int32_t *retval) int __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) { - char *bufp = NULL; + char *bufp = NULL; struct image_params *imgp; struct vnode_attr *vap; struct vnode_attr *origvap; int error; int is_64 = IS_64BIT_PROCESS(p); struct vfs_context context; - struct uthread *uthread; + struct uthread *uthread; + task_t old_task = current_task(); + task_t new_task = NULL; + boolean_t should_release_proc_ref = FALSE; + boolean_t exec_done = FALSE; + boolean_t in_vfexec = FALSE; + void *inherit = NULL; context.vc_thread = current_thread(); - context.vc_ucred = kauth_cred_proc_ref(p); /* XXX must NOT be kauth_cred_get() */ + context.vc_ucred = kauth_cred_proc_ref(p); /* XXX must NOT be kauth_cred_get() */ - /* Allocate a big chunk for locals instead of using stack since these + /* Allocate a big chunk for locals instead of using stack since these * structures a pretty big. */ MALLOC(bufp, char *, (sizeof(*imgp) + sizeof(*vap) + sizeof(*origvap)), M_TEMP, M_WAITOK | M_ZERO); @@ -3117,7 +3588,7 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) } vap = (struct vnode_attr *) (bufp + sizeof(*imgp)); origvap = (struct vnode_attr *) (bufp + sizeof(*imgp) + sizeof(*vap)); - + /* Initialize the common data in the image_params structure */ imgp->ip_user_fname = uap->fname; imgp->ip_user_argv = uap->argp; @@ -3125,14 +3596,10 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) imgp->ip_vattr = vap; imgp->ip_origvattr = origvap; imgp->ip_vfs_context = &context; - imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT : IMGPF_NONE) | ((p->p_flag & P_DISABLE_ASLR) ? IMGPF_DISABLE_ASLR : IMGPF_NONE); + imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT_ADDR : IMGPF_NONE) | ((p->p_flag & P_DISABLE_ASLR) ? IMGPF_DISABLE_ASLR : IMGPF_NONE); imgp->ip_seg = (is_64 ? UIO_USERSPACE64 : UIO_USERSPACE32); imgp->ip_mac_return = 0; - - uthread = get_bsdthread_info(current_thread()); - if (uthread->uu_flag & UT_VFORK) { - imgp->ip_flags |= IMGPF_VFORK_EXEC; - } + imgp->ip_cs_error = OS_REASON_NULL; #if CONFIG_MACF if (uap->mac_p != USER_ADDR_NULL) { @@ -3143,31 +3610,122 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) } } #endif + uthread = get_bsdthread_info(current_thread()); + if (uthread->uu_flag & UT_VFORK) { + imgp->ip_flags |= IMGPF_VFORK_EXEC; + in_vfexec = TRUE; + } else { + imgp->ip_flags |= IMGPF_EXEC; - error = exec_activate_image(imgp); + /* + * For execve case, create a new task and thread + * which points to current_proc. The current_proc will point + * to the new task after image activation and proc ref drain. + * + * proc (current_proc) <----- old_task (current_task) + * ^ | ^ + * | | | + * | ---------------------------------- + * | + * --------- new_task (task marked as TF_EXEC_COPY) + * + * After image activation, the proc will point to the new task + * and would look like following. + * + * proc (current_proc) <----- old_task (current_task, marked as TPF_DID_EXEC) + * ^ | + * | | + * | ----------> new_task + * | | + * ----------------- + * + * During exec any transition from new_task -> proc is fine, but don't allow + * transition from proc->task, since it will modify old_task. + */ + imgp->ip_new_thread = fork_create_child(old_task, + NULL, + p, + FALSE, + p->p_flag & P_LP64, + task_get_64bit_data(old_task), + TRUE); + /* task and thread ref returned by fork_create_child */ + if (imgp->ip_new_thread == NULL) { + error = ENOMEM; + goto exit_with_error; + } - kauth_cred_unref(&context.vc_ucred); - - /* Image not claimed by any activator? */ - if (error == -1) + new_task = get_threadtask(imgp->ip_new_thread); + context.vc_thread = imgp->ip_new_thread; + } + + error = exec_activate_image(imgp); + /* thread and task ref returned for vfexec case */ + + if (imgp->ip_new_thread != NULL) { + /* + * task reference might be returned by exec_activate_image + * for vfexec. + */ + new_task = get_threadtask(imgp->ip_new_thread); + } + + if (!error && !in_vfexec) { + p = proc_exec_switch_task(p, old_task, new_task, imgp->ip_new_thread); + /* proc ref returned */ + should_release_proc_ref = TRUE; + + /* + * Need to transfer pending watch port boosts to the new task while still making + * sure that the old task remains in the importance linkage. Create an importance + * linkage from old task to new task, then switch the task importance base + * of old task and new task. After the switch the port watch boost will be + * boosting the new task and new task will be donating importance to old task. + */ + inherit = ipc_importance_exec_switch_task(old_task, new_task); + } + + kauth_cred_unref(&context.vc_ucred); + + /* Image not claimed by any activator? */ + if (error == -1) { error = ENOEXEC; + } + + if (!error) { + exec_done = TRUE; + assert(imgp->ip_new_thread != NULL); - if (error == 0) { exec_resettextvp(p, imgp); error = check_for_signature(p, imgp); - } - if (imgp->ip_vp != NULLVP) + } + + /* flag exec has occurred, notify only if it has not failed due to FP Key error */ + if (exec_done && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) { + proc_knote(p, NOTE_EXEC); + } + + if (imgp->ip_vp != NULLVP) { vnode_put(imgp->ip_vp); - if (imgp->ip_scriptvp != NULLVP) + } + if (imgp->ip_scriptvp != NULLVP) { vnode_put(imgp->ip_scriptvp); - if (imgp->ip_strings) + } + if (imgp->ip_strings) { execargs_free(imgp); + } #if CONFIG_MACF - if (imgp->ip_execlabelp) + if (imgp->ip_execlabelp) { mac_cred_label_free(imgp->ip_execlabelp); - if (imgp->ip_scriptlabelp) + } + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } #endif + if (imgp->ip_cs_error != OS_REASON_NULL) { + os_reason_free(imgp->ip_cs_error); + imgp->ip_cs_error = OS_REASON_NULL; + } if (!error) { /* @@ -3177,45 +3735,126 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) * aren't loaded until subsequent calls (including exec_resettextvp). */ error = proc_transstart(p, 0, 0); - - if (!error) { - task_bank_init(p->task); - proc_transend(p, 0); - } } if (!error) { + task_bank_init(new_task); + proc_transend(p, 0); + +#if __arm64__ + proc_legacy_footprint(p, new_task, __FUNCTION__); +#endif /* __arm64__ */ + /* Sever any extant thread affinity */ thread_affinity_exec(current_thread()); - thread_t main_thread = (imgp->ip_new_thread != NULL) ? imgp->ip_new_thread : current_thread(); + /* Inherit task role from old task to new task for exec */ + if (!in_vfexec) { + proc_inherit_task_role(new_task, old_task); + } - task_set_main_thread_qos(p->task, main_thread); + thread_t main_thread = imgp->ip_new_thread; - /* reset task importance */ - task_importance_reset(p->task); + task_set_main_thread_qos(new_task, main_thread); + +#if CONFIG_MACF + /* + * Processes with the MAP_JIT entitlement are permitted to have + * a jumbo-size map. + */ + if (mac_proc_check_map_anon(p, 0, 0, 0, MAP_JIT, NULL) == 0) { + vm_map_set_jumbo(get_task_map(new_task)); + } +#endif /* CONFIG_MACF */ - DTRACE_PROC(exec__success); + if (vm_darkwake_mode == TRUE) { + /* + * This process is being launched when the system + * is in darkwake. So mark it specially. This will + * cause all its pages to be entered in the background Q. + */ + task_set_darkwake_mode(new_task, vm_darkwake_mode); + } #if CONFIG_DTRACE - if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) + dtrace_thread_didexec(imgp->ip_new_thread); + + if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) { (*dtrace_proc_waitfor_hook)(p); + } #endif - if (imgp->ip_flags & IMGPF_VFORK_EXEC) { +#if CONFIG_AUDIT + if (!error && AUDIT_ENABLED() && p) { + /* Add the CDHash of the new process to the audit record */ + uint8_t *cdhash = cs_get_cdhash(p); + if (cdhash) { + AUDIT_ARG(data, cdhash, sizeof(uint8_t), CS_CDHASH_LEN); + } + } +#endif + + if (in_vfexec) { vfork_return(p, retval, p->p_pid); - proc_clear_return_wait(p, imgp->ip_new_thread); } } else { DTRACE_PROC1(exec__failure, int, error); } exit_with_error: + + /* + * clear bsd_info from old task if it did exec. + */ + if (task_did_exec(old_task)) { + set_bsdtask_info(old_task, NULL); + } + + /* clear bsd_info from new task and terminate it if exec failed */ + if (new_task != NULL && task_is_exec_copy(new_task)) { + set_bsdtask_info(new_task, NULL); + task_terminate_internal(new_task); + } + + if (imgp != NULL) { + /* + * Do not terminate the current task, if proc_exec_switch_task did not + * switch the tasks, terminating the current task without the switch would + * result in loosing the SIGKILL status. + */ + if (task_did_exec(old_task)) { + /* Terminate the current task, since exec will start in new task */ + task_terminate_internal(old_task); + } + + /* Release the thread ref returned by fork_create_child */ + if (imgp->ip_new_thread) { + /* wake up the new exec thread */ + task_clear_return_wait(get_threadtask(imgp->ip_new_thread)); + thread_deallocate(imgp->ip_new_thread); + imgp->ip_new_thread = NULL; + } + } + + /* Release the ref returned by fork_create_child */ + if (new_task) { + task_deallocate(new_task); + new_task = NULL; + } + + if (should_release_proc_ref) { + proc_rele(p); + } + if (bufp != NULL) { FREE(bufp, M_TEMP); } - - return(error); + + if (inherit != NULL) { + ipc_importance_release(inherit); + } + + return error; } @@ -3245,11 +3884,11 @@ copyinptr(user_addr_t froma, user_addr_t *toptr, int ptr_size) unsigned int i; error = copyin(froma, &i, 4); - *toptr = CAST_USER_ADDR_T(i); /* SAFE */ + *toptr = CAST_USER_ADDR_T(i); /* SAFE */ } else { error = copyin(froma, toptr, 8); } - return (error); + return error; } @@ -3274,13 +3913,13 @@ copyoutptr(user_addr_t ua, user_addr_t ptr, int ptr_size) if (ptr_size == 4) { /* 64 bit value containing 32 bit address */ - unsigned int i = CAST_DOWN_EXPLICIT(unsigned int,ua); /* SAFE */ + unsigned int i = CAST_DOWN_EXPLICIT(unsigned int, ua); /* SAFE */ error = copyout(&i, ptr, 4); } else { error = copyout(&ua, ptr, 8); } - return (error); + return error; } @@ -3362,25 +4001,25 @@ static int exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) { proc_t p = vfs_context_proc(imgp->ip_vfs_context); - int ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; - int ptr_area_size; + int ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; + int ptr_area_size; void *ptr_buffer_start, *ptr_buffer; int string_size; - user_addr_t string_area; /* *argv[], *env[] */ - user_addr_t ptr_area; /* argv[], env[], applev[] */ - user_addr_t argc_area; /* argc */ - user_addr_t stack; + user_addr_t string_area; /* *argv[], *env[] */ + user_addr_t ptr_area; /* argv[], env[], applev[] */ + user_addr_t argc_area; /* argc */ + user_addr_t stack; int error; unsigned i; struct copyout_desc { - char *start_string; - int count; + char *start_string; + int count; #if CONFIG_DTRACE - user_addr_t *dtrace_cookie; + user_addr_t *dtrace_cookie; #endif - boolean_t null_term; + boolean_t null_term; } descriptors[] = { { .start_string = imgp->ip_startargv, @@ -3436,8 +4075,7 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Need room for one pointer for each string, plus * one for the NULLs terminating the argv, envv, and apple areas. */ - ptr_area_size = (imgp->ip_argc + imgp->ip_envc + imgp->ip_applec + 3) * - ptr_size; + ptr_area_size = (imgp->ip_argc + imgp->ip_envc + imgp->ip_applec + 3) * ptr_size; stack -= ptr_area_size; ptr_area = stack; @@ -3469,11 +4107,12 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Copy out the entire strings area. */ error = copyout(imgp->ip_strings, string_area, - string_size); - if (error) + string_size); + if (error) { goto bad; + } - for (i = 0; i < sizeof(descriptors)/sizeof(descriptors[0]); i++) { + for (i = 0; i < sizeof(descriptors) / sizeof(descriptors[0]); i++) { char *cur_string = descriptors[i].start_string; int j; @@ -3491,14 +4130,14 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) */ for (j = 0; j < descriptors[i].count; j++) { user_addr_t cur_address = string_area + (cur_string - imgp->ip_strings); - + /* Copy out the pointer to the current string. Alignment has been verified */ if (ptr_size == 8) { *(uint64_t *)ptr_buffer = (uint64_t)cur_address; } else { *(uint32_t *)ptr_buffer = (uint32_t)cur_address; } - + ptr_buffer = (void *)((uintptr_t)ptr_buffer + ptr_size); cur_string += strlen(cur_string) + 1; /* Only a NUL between strings in the same area */ } @@ -3509,7 +4148,7 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) } else { *(uint32_t *)ptr_buffer = 0; } - + ptr_buffer = (void *)((uintptr_t)ptr_buffer + ptr_size); } } @@ -3518,17 +4157,19 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Copy out all our pointer arrays in bulk. */ error = copyout(ptr_buffer_start, ptr_area, - ptr_area_size); - if (error) + ptr_area_size); + if (error) { goto bad; + } /* argc (int32, stored in a ptr_size area) */ error = copyoutptr((user_addr_t)imgp->ip_argc, argc_area, ptr_size); - if (error) + if (error) { goto bad; + } bad: - return(error); + return error; } @@ -3563,14 +4204,14 @@ static int exec_extract_strings(struct image_params *imgp) { int error = 0; - int ptr_size = (imgp->ip_flags & IMGPF_WAS_64BIT) ? 8 : 4; - int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; - user_addr_t argv = imgp->ip_user_argv; - user_addr_t envv = imgp->ip_user_envv; + int ptr_size = (imgp->ip_flags & IMGPF_WAS_64BIT_ADDR) ? 8 : 4; + int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; + user_addr_t argv = imgp->ip_user_argv; + user_addr_t envv = imgp->ip_user_envv; /* * Adjust space reserved for the path name by however much padding it - * needs. Doing this here since we didn't know if this would be a 32- + * needs. Doing this here since we didn't know if this would be a 32- * or 64-bit process back in exec_save_path. */ while (imgp->ip_strspace % new_ptr_size != 0) { @@ -3585,8 +4226,8 @@ exec_extract_strings(struct image_params *imgp) imgp->ip_startargv = imgp->ip_strendp; imgp->ip_argc = 0; - if((imgp->ip_flags & IMGPF_INTERPRET) != 0) { - user_addr_t arg; + if ((imgp->ip_flags & IMGPF_INTERPRET) != 0) { + user_addr_t arg; char *argstart, *ch; /* First, the arguments in the "#!" string are tokenized and extracted. */ @@ -3617,8 +4258,9 @@ exec_extract_strings(struct image_params *imgp) } /* Error-check, regardless of whether this is the last interpreter arg or not */ - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -3635,8 +4277,9 @@ exec_extract_strings(struct image_params *imgp) * to locate their script arguments. */ error = copyinptr(argv, &arg, ptr_size); - if (error) + if (error) { goto bad; + } if (arg != 0LL) { argv += ptr_size; /* consume without using */ } @@ -3649,9 +4292,10 @@ exec_extract_strings(struct image_params *imgp) } else { error = exec_add_user_string(imgp, imgp->ip_user_fname, imgp->ip_seg, TRUE); } - - if (error) + + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -3661,11 +4305,12 @@ exec_extract_strings(struct image_params *imgp) } while (argv != 0LL) { - user_addr_t arg; + user_addr_t arg; error = copyinptr(argv, &arg, ptr_size); - if (error) + if (error) { goto bad; + } if (arg == 0LL) { break; @@ -3674,18 +4319,19 @@ exec_extract_strings(struct image_params *imgp) argv += ptr_size; /* - * av[n...] = arg[n] - */ + * av[n...] = arg[n] + */ error = exec_add_user_string(imgp, arg, imgp->ip_seg, TRUE); - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; } imgp->ip_argspace -= new_ptr_size; /* to hold argv[] entry */ imgp->ip_argc++; - } + } /* Save space for argv[] NULL terminator */ if (imgp->ip_argspace < new_ptr_size) { @@ -3693,29 +4339,31 @@ exec_extract_strings(struct image_params *imgp) goto bad; } imgp->ip_argspace -= new_ptr_size; - + /* Note where the args ends and env begins. */ imgp->ip_endargv = imgp->ip_strendp; imgp->ip_envc = 0; /* Now, get the environment */ while (envv != 0LL) { - user_addr_t env; + user_addr_t env; error = copyinptr(envv, &env, ptr_size); - if (error) + if (error) { goto bad; + } envv += ptr_size; if (env == 0LL) { break; } /* - * av[n...] = env[n] - */ + * av[n...] = env[n] + */ error = exec_add_user_string(imgp, env, imgp->ip_seg, TRUE); - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -3741,7 +4389,7 @@ exec_extract_strings(struct image_params *imgp) imgp->ip_strspace--; imgp->ip_argspace--; } - + /* Note where the envv ends and applev begins. */ imgp->ip_endenvv = imgp->ip_strendp; @@ -3754,62 +4402,27 @@ bad: return error; } -static char * -random_hex_str(char *str, int len, boolean_t embedNUL) -{ - uint64_t low, high, value; - int idx; - char digit; - - /* A 64-bit value will only take 16 characters, plus '0x' and NULL. */ - if (len > 19) - len = 19; - - /* We need enough room for at least 1 digit */ - if (len < 4) - return (NULL); - - low = random(); - high = random(); - value = high << 32 | low; - - if (embedNUL) { - /* - * Zero a byte to protect against C string vulnerabilities - * e.g. for userland __stack_chk_guard. - */ - value &= ~(0xffull << 8); - } - - str[0] = '0'; - str[1] = 'x'; - for (idx = 2; idx < len - 1; idx++) { - digit = value & 0xf; - value = value >> 4; - if (digit < 10) - str[idx] = '0' + digit; - else - str[idx] = 'a' + (digit - 10); - } - str[idx] = '\0'; - return (str); -} - /* * Libc has an 8-element array set up for stack guard values. It only fills * in one of those entries, and both gcc and llvm seem to use only a single * 8-byte guard. Until somebody needs more than an 8-byte guard value, don't * do the work to construct them. */ -#define GUARD_VALUES 1 -#define GUARD_KEY "stack_guard=" +#define GUARD_VALUES 1 +#define GUARD_KEY "stack_guard=" /* * System malloc needs some entropy when it is initialized. */ -#define ENTROPY_VALUES 2 +#define ENTROPY_VALUES 2 #define ENTROPY_KEY "malloc_entropy=" +/* + * libplatform needs a random pointer-obfuscation value when it is initialized. + */ +#define PTR_MUNGE_VALUES 1 +#define PTR_MUNGE_KEY "ptr_munge=" + /* * System malloc engages nanozone for UIAPP. */ @@ -3818,49 +4431,87 @@ random_hex_str(char *str, int len, boolean_t embedNUL) #define PFZ_KEY "pfz=" extern user32_addr_t commpage_text32_location; extern user64_addr_t commpage_text64_location; + +#define MAIN_STACK_VALUES 4 +#define MAIN_STACK_KEY "main_stack=" + +#define FSID_KEY "executable_file=" +#define DYLD_FSID_KEY "dyld_file=" +#define CDHASH_KEY "executable_cdhash=" + +#define FSID_MAX_STRING "0x1234567890abcdef,0x1234567890abcdef" + +#define HEX_STR_LEN 18 // 64-bit hex value "0x0123456701234567" + +static int +exec_add_entropy_key(struct image_params *imgp, + const char *key, + int values, + boolean_t embedNUL) +{ + const int limit = 8; + uint64_t entropy[limit]; + char str[strlen(key) + (HEX_STR_LEN + 1) * limit + 1]; + if (values > limit) { + values = limit; + } + + read_random(entropy, sizeof(entropy[0]) * values); + + if (embedNUL) { + entropy[0] &= ~(0xffull << 8); + } + + int len = snprintf(str, sizeof(str), "%s0x%llx", key, entropy[0]); + int remaining = sizeof(str) - len; + for (int i = 1; i < values && remaining > 0; ++i) { + int start = sizeof(str) - remaining; + len = snprintf(&str[start], remaining, ",0x%llx", entropy[i]); + remaining -= len; + } + + return exec_add_user_string(imgp, CAST_USER_ADDR_T(str), UIO_SYSSPACE, FALSE); +} + /* * Build up the contents of the apple[] string vector */ static int -exec_add_apple_strings(struct image_params *imgp) +exec_add_apple_strings(struct image_params *imgp, + const load_result_t *load_result) { - int i, error; - int new_ptr_size=4; - char guard[19]; - char guard_vec[strlen(GUARD_KEY) + 19 * GUARD_VALUES + 1]; - - char entropy[19]; - char entropy_vec[strlen(ENTROPY_KEY) + 19 * ENTROPY_VALUES + 1]; - - char pfz_string[strlen(PFZ_KEY) + 16 + 4 +1]; - - if( imgp->ip_flags & IMGPF_IS_64BIT) { - new_ptr_size = 8; - snprintf(pfz_string, sizeof(pfz_string),PFZ_KEY "0x%llx",commpage_text64_location); - } else { - snprintf(pfz_string, sizeof(pfz_string),PFZ_KEY "0x%x",commpage_text32_location); - } + int error; + int img_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; /* exec_save_path stored the first string */ imgp->ip_applec = 1; /* adding the pfz string */ - error = exec_add_user_string(imgp, CAST_USER_ADDR_T(pfz_string),UIO_SYSSPACE,FALSE); - if(error) - goto bad; - imgp->ip_applec++; + { + char pfz_string[strlen(PFZ_KEY) + HEX_STR_LEN + 1]; + + if (img_ptr_size == 8) { + snprintf(pfz_string, sizeof(pfz_string), PFZ_KEY "0x%llx", commpage_text64_location); + } else { + snprintf(pfz_string, sizeof(pfz_string), PFZ_KEY "0x%x", commpage_text32_location); + } + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(pfz_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } /* adding the NANO_ENGAGE_KEY key */ if (imgp->ip_px_sa) { int proc_flags = (((struct _posix_spawnattr *) imgp->ip_px_sa)->psa_flags); if ((proc_flags & _POSIX_SPAWN_NANO_ALLOCATOR) == _POSIX_SPAWN_NANO_ALLOCATOR) { - char uiapp_string[strlen(NANO_ENGAGE_KEY) + 1]; - - snprintf(uiapp_string, sizeof(uiapp_string), NANO_ENGAGE_KEY); - error = exec_add_user_string(imgp, CAST_USER_ADDR_T(uiapp_string),UIO_SYSSPACE,FALSE); - if (error) + const char *nano_string = NANO_ENGAGE_KEY; + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(nano_string), UIO_SYSSPACE, FALSE); + if (error) { goto bad; + } imgp->ip_applec++; } } @@ -3872,37 +4523,97 @@ exec_add_apple_strings(struct image_params *imgp) * (The first random string always contains an embedded NUL so that * __stack_chk_guard also protects against C string vulnerabilities) */ - (void)strlcpy(guard_vec, GUARD_KEY, sizeof (guard_vec)); - for (i = 0; i < GUARD_VALUES; i++) { - random_hex_str(guard, sizeof (guard), i == 0); - if (i) - (void)strlcat(guard_vec, ",", sizeof (guard_vec)); - (void)strlcat(guard_vec, guard, sizeof (guard_vec)); + error = exec_add_entropy_key(imgp, GUARD_KEY, GUARD_VALUES, TRUE); + if (error) { + goto bad; } + imgp->ip_applec++; - error = exec_add_user_string(imgp, CAST_USER_ADDR_T(guard_vec), UIO_SYSSPACE, FALSE); - if (error) + /* + * Supply libc with entropy for system malloc. + */ + error = exec_add_entropy_key(imgp, ENTROPY_KEY, ENTROPY_VALUES, FALSE); + if (error) { goto bad; + } imgp->ip_applec++; /* - * Supply libc with entropy for system malloc. + * Supply libpthread & libplatform with a random value to use for pointer + * obfuscation. */ - (void)strlcpy(entropy_vec, ENTROPY_KEY, sizeof(entropy_vec)); - for (i = 0; i < ENTROPY_VALUES; i++) { - random_hex_str(entropy, sizeof (entropy), FALSE); - if (i) - (void)strlcat(entropy_vec, ",", sizeof (entropy_vec)); - (void)strlcat(entropy_vec, entropy, sizeof (entropy_vec)); - } - - error = exec_add_user_string(imgp, CAST_USER_ADDR_T(entropy_vec), UIO_SYSSPACE, FALSE); - if (error) + error = exec_add_entropy_key(imgp, PTR_MUNGE_KEY, PTR_MUNGE_VALUES, FALSE); + if (error) { goto bad; + } imgp->ip_applec++; + /* + * Add MAIN_STACK_KEY: Supplies the address and size of the main thread's + * stack if it was allocated by the kernel. + * + * The guard page is not included in this stack size as libpthread + * expects to add it back in after receiving this value. + */ + if (load_result->unixproc) { + char stack_string[strlen(MAIN_STACK_KEY) + (HEX_STR_LEN + 1) * MAIN_STACK_VALUES + 1]; + snprintf(stack_string, sizeof(stack_string), + MAIN_STACK_KEY "0x%llx,0x%llx,0x%llx,0x%llx", + (uint64_t)load_result->user_stack, + (uint64_t)load_result->user_stack_size, + (uint64_t)load_result->user_stack_alloc, + (uint64_t)load_result->user_stack_alloc_size); + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(stack_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } + + if (imgp->ip_vattr) { + uint64_t fsid = vnode_get_va_fsid(imgp->ip_vattr); + uint64_t fsobjid = imgp->ip_vattr->va_fileid; + + char fsid_string[strlen(FSID_KEY) + strlen(FSID_MAX_STRING) + 1]; + snprintf(fsid_string, sizeof(fsid_string), + FSID_KEY "0x%llx,0x%llx", fsid, fsobjid); + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(fsid_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } + + if (imgp->ip_dyld_fsid || imgp->ip_dyld_fsobjid) { + char fsid_string[strlen(DYLD_FSID_KEY) + strlen(FSID_MAX_STRING) + 1]; + snprintf(fsid_string, sizeof(fsid_string), + DYLD_FSID_KEY "0x%llx,0x%llx", imgp->ip_dyld_fsid, imgp->ip_dyld_fsobjid); + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(fsid_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } + + uint8_t cdhash[SHA1_RESULTLEN]; + int cdhash_errror = ubc_cs_getcdhash(imgp->ip_vp, imgp->ip_arch_offset, cdhash); + if (cdhash_errror == 0) { + char hash_string[strlen(CDHASH_KEY) + 2 * SHA1_RESULTLEN + 1]; + strncpy(hash_string, CDHASH_KEY, sizeof(hash_string)); + char *p = hash_string + sizeof(CDHASH_KEY) - 1; + for (int i = 0; i < SHA1_RESULTLEN; i++) { + snprintf(p, 3, "%02x", (int) cdhash[i]); + p += 2; + } + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(hash_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } + /* Align the tail of the combined applev area */ - while (imgp->ip_strspace % new_ptr_size != 0) { + while (imgp->ip_strspace % img_ptr_size != 0) { *imgp->ip_strendp++ = '\0'; imgp->ip_strspace--; } @@ -3911,7 +4622,7 @@ bad: return error; } -#define unix_stack_size(p) (p->p_rlimit[RLIMIT_STACK].rlim_cur) +#define unix_stack_size(p) (p->p_rlimit[RLIMIT_STACK].rlim_cur) /* * exec_check_permissions @@ -3939,59 +4650,69 @@ exec_check_permissions(struct image_params *imgp) kauth_action_t action; /* Only allow execution of regular files */ - if (!vnode_isreg(vp)) - return (EACCES); - + if (!vnode_isreg(vp)) { + return EACCES; + } + /* Get the file attributes that we will be using here and elsewhere */ VATTR_INIT(vap); VATTR_WANTED(vap, va_uid); VATTR_WANTED(vap, va_gid); VATTR_WANTED(vap, va_mode); VATTR_WANTED(vap, va_fsid); + VATTR_WANTED(vap, va_fsid64); VATTR_WANTED(vap, va_fileid); VATTR_WANTED(vap, va_data_size); - if ((error = vnode_getattr(vp, vap, imgp->ip_vfs_context)) != 0) - return (error); + if ((error = vnode_getattr(vp, vap, imgp->ip_vfs_context)) != 0) { + return error; + } /* * Ensure that at least one execute bit is on - otherwise root * will always succeed, and we don't want to happen unless the * file really is executable. */ - if (!vfs_authopaque(vnode_mount(vp)) && ((vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) - return (EACCES); + if (!vfs_authopaque(vnode_mount(vp)) && ((vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) { + return EACCES; + } /* Disallow zero length files */ - if (vap->va_data_size == 0) - return (ENOEXEC); + if (vap->va_data_size == 0) { + return ENOEXEC; + } imgp->ip_arch_offset = (user_size_t)0; imgp->ip_arch_size = vap->va_data_size; /* Disable setuid-ness for traced programs or if MNT_NOSUID */ - if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_lflag & P_LTRACED)) + if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_lflag & P_LTRACED)) { vap->va_mode &= ~(VSUID | VSGID); + } /* * Disable _POSIX_SPAWN_ALLOW_DATA_EXEC and _POSIX_SPAWN_DISABLE_ASLR * flags for setuid/setgid binaries. */ - if (vap->va_mode & (VSUID | VSGID)) + if (vap->va_mode & (VSUID | VSGID)) { imgp->ip_flags &= ~(IMGPF_ALLOW_DATA_EXEC | IMGPF_DISABLE_ASLR); + } #if CONFIG_MACF error = mac_vnode_check_exec(imgp->ip_vfs_context, vp, imgp); - if (error) - return (error); + if (error) { + return error; + } #endif - /* Check for execute permission */ - action = KAUTH_VNODE_EXECUTE; - /* Traced images must also be readable */ - if (p->p_lflag & P_LTRACED) - action |= KAUTH_VNODE_READ_DATA; - if ((error = vnode_authorize(vp, NULL, action, imgp->ip_vfs_context)) != 0) - return (error); + /* Check for execute permission */ + action = KAUTH_VNODE_EXECUTE; + /* Traced images must also be readable */ + if (p->p_lflag & P_LTRACED) { + action |= KAUTH_VNODE_READ_DATA; + } + if ((error = vnode_authorize(vp, NULL, action, imgp->ip_vfs_context)) != 0) { + return error; + } #if 0 /* Don't let it run if anyone had it open for writing */ @@ -3999,7 +4720,7 @@ exec_check_permissions(struct image_params *imgp) if (vp->v_writecount) { panic("going to return ETXTBSY %x", vp); vnode_unlock(vp); - return (ETXTBSY); + return ETXTBSY; } vnode_unlock(vp); #endif @@ -4007,7 +4728,7 @@ exec_check_permissions(struct image_params *imgp) /* XXX May want to indicate to underlying FS that vnode is open */ - return (error); + return error; } @@ -4040,15 +4761,17 @@ exec_check_permissions(struct image_params *imgp) static int exec_handle_sugid(struct image_params *imgp) { - kauth_cred_t cred = vfs_context_ucred(imgp->ip_vfs_context); - proc_t p = vfs_context_proc(imgp->ip_vfs_context); - int i; - int leave_sugid_clear = 0; - int mac_reset_ipc = 0; - int error = 0; + proc_t p = vfs_context_proc(imgp->ip_vfs_context); + kauth_cred_t cred = vfs_context_ucred(imgp->ip_vfs_context); + kauth_cred_t my_cred, my_new_cred; + int i; + int leave_sugid_clear = 0; + int mac_reset_ipc = 0; + int error = 0; + task_t task = NULL; #if CONFIG_MACF - int mac_transition, disjoint_cred = 0; - int label_update_return = 0; + int mac_transition, disjoint_cred = 0; + int label_update_return = 0; /* * Determine whether a call to update the MAC label will result in the @@ -4060,14 +4783,14 @@ exec_handle_sugid(struct image_params *imgp) * slow down the exec fast path for normal binaries. */ mac_transition = mac_cred_check_label_update_execve( - imgp->ip_vfs_context, - imgp->ip_vp, - imgp->ip_arch_offset, - imgp->ip_scriptvp, - imgp->ip_scriptlabelp, - imgp->ip_execlabelp, - p, - imgp->ip_px_smpx); + imgp->ip_vfs_context, + imgp->ip_vp, + imgp->ip_arch_offset, + imgp->ip_scriptvp, + imgp->ip_scriptlabelp, + imgp->ip_execlabelp, + p, + imgp->ip_px_smpx); #endif OSBitAndAtomic(~((uint32_t)P_SUGID), &p->p_flag); @@ -4087,16 +4810,16 @@ exec_handle_sugid(struct image_params *imgp) * such a call. */ if (((imgp->ip_origvattr->va_mode & VSUID) != 0 && - kauth_cred_getuid(cred) != imgp->ip_origvattr->va_uid) || + kauth_cred_getuid(cred) != imgp->ip_origvattr->va_uid) || ((imgp->ip_origvattr->va_mode & VSGID) != 0 && - ((kauth_cred_ismember_gid(cred, imgp->ip_origvattr->va_gid, &leave_sugid_clear) || !leave_sugid_clear) || - (kauth_cred_getgid(cred) != imgp->ip_origvattr->va_gid)))) { - + ((kauth_cred_ismember_gid(cred, imgp->ip_origvattr->va_gid, &leave_sugid_clear) || !leave_sugid_clear) || + (kauth_cred_getgid(cred) != imgp->ip_origvattr->va_gid)))) { #if CONFIG_MACF /* label for MAC transition and neither VSUID nor VSGID */ handle_mac_transition: #endif +#if !SECURE_KERNEL /* * Replace the credential with a copy of itself if euid or * egid change. @@ -4112,26 +4835,78 @@ handle_mac_transition: * membership resolution, then dropping their * effective privilege to that of the desired * final credential state. + * + * Modifications to p_ucred must be guarded using the + * proc's ucred lock. This prevents others from accessing + * a garbage credential. */ - if (imgp->ip_origvattr->va_mode & VSUID) { - p->p_ucred = kauth_cred_setresuid(p->p_ucred, KAUTH_UID_NONE, imgp->ip_origvattr->va_uid, imgp->ip_origvattr->va_uid, KAUTH_UID_NONE); + while (imgp->ip_origvattr->va_mode & VSUID) { + my_cred = kauth_cred_proc_ref(p); + my_new_cred = kauth_cred_setresuid(my_cred, KAUTH_UID_NONE, imgp->ip_origvattr->va_uid, imgp->ip_origvattr->va_uid, KAUTH_UID_NONE); + + if (my_new_cred == my_cred) { + kauth_cred_unref(&my_cred); + break; + } + /* update cred on proc */ + proc_ucred_lock(p); + + if (p->p_ucred != my_cred) { + proc_ucred_unlock(p); + kauth_cred_unref(&my_new_cred); + continue; + } + + /* donate cred reference on my_new_cred to p->p_ucred */ + p->p_ucred = my_new_cred; PROC_UPDATE_CREDS_ONPROC(p); + proc_ucred_unlock(p); + + /* drop additional reference that was taken on the previous cred */ + kauth_cred_unref(&my_cred); + + break; } - if (imgp->ip_origvattr->va_mode & VSGID) { - p->p_ucred = kauth_cred_setresgid(p->p_ucred, KAUTH_GID_NONE, imgp->ip_origvattr->va_gid, imgp->ip_origvattr->va_gid); + + while (imgp->ip_origvattr->va_mode & VSGID) { + my_cred = kauth_cred_proc_ref(p); + my_new_cred = kauth_cred_setresgid(my_cred, KAUTH_GID_NONE, imgp->ip_origvattr->va_gid, imgp->ip_origvattr->va_gid); + + if (my_new_cred == my_cred) { + kauth_cred_unref(&my_cred); + break; + } + /* update cred on proc */ + proc_ucred_lock(p); + + if (p->p_ucred != my_cred) { + proc_ucred_unlock(p); + kauth_cred_unref(&my_new_cred); + continue; + } + + /* donate cred reference on my_new_cred to p->p_ucred */ + p->p_ucred = my_new_cred; PROC_UPDATE_CREDS_ONPROC(p); + proc_ucred_unlock(p); + + /* drop additional reference that was taken on the previous cred */ + kauth_cred_unref(&my_cred); + + break; } +#endif /* !SECURE_KERNEL */ #if CONFIG_MACF - /* + /* * If a policy has indicated that it will transition the label, * before making the call into the MAC policies, get a new * duplicate credential, so they can modify it without * modifying any others sharing it. */ - if (mac_transition) { + if (mac_transition) { /* * This hook may generate upcalls that require * importance donation from the kernel. @@ -4140,16 +4915,16 @@ handle_mac_transition: thread_t thread = current_thread(); thread_enable_send_importance(thread, TRUE); kauth_proc_label_update_execve(p, - imgp->ip_vfs_context, - imgp->ip_vp, - imgp->ip_arch_offset, - imgp->ip_scriptvp, - imgp->ip_scriptlabelp, - imgp->ip_execlabelp, - &imgp->ip_csflags, - imgp->ip_px_smpx, - &disjoint_cred, /* will be non zero if disjoint */ - &label_update_return); + imgp->ip_vfs_context, + imgp->ip_vp, + imgp->ip_arch_offset, + imgp->ip_scriptvp, + imgp->ip_scriptlabelp, + imgp->ip_execlabelp, + &imgp->ip_csflags, + imgp->ip_px_smpx, + &disjoint_cred, /* will be non zero if disjoint */ + &label_update_return); thread_enable_send_importance(thread, FALSE); if (disjoint_cred) { @@ -4164,13 +4939,13 @@ handle_mac_transition: */ leave_sugid_clear = 0; } - + imgp->ip_mac_return = label_update_return; } - + mac_reset_ipc = mac_proc_check_inherit_ipc_ports(p, p->p_textvp, p->p_textoff, imgp->ip_vp, imgp->ip_arch_offset, imgp->ip_scriptvp); -#endif /* CONFIG_MACF */ +#endif /* CONFIG_MACF */ /* * If 'leave_sugid_clear' is non-zero, then we passed the @@ -4178,7 +4953,7 @@ handle_mac_transition: * the previous cred was a member of the VSGID group, but * that it was not the default at the time of the execve, * and that the post-labelling credential was not disjoint. - * So we don't set the P_SUGID or reset mach ports and fds + * So we don't set the P_SUGID or reset mach ports and fds * on the basis of simply running this code. */ if (mac_reset_ipc || !leave_sugid_clear) { @@ -4188,9 +4963,10 @@ handle_mac_transition: * a setuid exec to be able to access/control the * task/thread after. */ - ipc_task_reset(p->task); + ipc_task_reset((imgp->ip_new_thread != NULL) ? + get_threadtask(imgp->ip_new_thread) : p->task); ipc_thread_reset((imgp->ip_new_thread != NULL) ? - imgp->ip_new_thread : current_thread()); + imgp->ip_new_thread : current_thread()); } if (!leave_sugid_clear) { @@ -4208,17 +4984,17 @@ handle_mac_transition: * to libc. */ for (i = 0; i < 3; i++) { - - if (p->p_fd->fd_ofiles[i] != NULL) + if (p->p_fd->fd_ofiles[i] != NULL) { continue; + } /* * Do the kernel equivalent of * - * if i == 0 - * (void) open("/dev/null", O_RDONLY); - * else - * (void) open("/dev/null", O_WRONLY); + * if i == 0 + * (void) open("/dev/null", O_RDONLY); + * else + * (void) open("/dev/null", O_WRONLY); */ struct fileproc *fp; @@ -4226,17 +5002,20 @@ handle_mac_transition: int flag; struct nameidata *ndp = NULL; - if (i == 0) + if (i == 0) { flag = FREAD; - else + } else { flag = FWRITE; + } if ((error = falloc(p, - &fp, &indx, imgp->ip_vfs_context)) != 0) + &fp, &indx, imgp->ip_vfs_context)) != 0) { continue; + } MALLOC(ndp, struct nameidata *, sizeof(*ndp), M_TEMP, M_WAITOK | M_ZERO); if (ndp == NULL) { + fp_free(p, indx, fp); error = ENOMEM; break; } @@ -4247,6 +5026,7 @@ handle_mac_transition: if ((error = vn_open(ndp, flag, 0)) != 0) { fp_free(p, indx, fp); + FREE(ndp, M_TEMP); break; } @@ -4282,21 +5062,57 @@ handle_mac_transition: } } -#endif /* CONFIG_MACF */ +#endif /* CONFIG_MACF */ /* * Implement the semantic where the effective user and group become * the saved user and group in exec'ed programs. + * + * Modifications to p_ucred must be guarded using the + * proc's ucred lock. This prevents others from accessing + * a garbage credential. */ - p->p_ucred = kauth_cred_setsvuidgid(p->p_ucred, kauth_cred_getuid(p->p_ucred), kauth_cred_getgid(p->p_ucred)); - /* update cred on proc */ - PROC_UPDATE_CREDS_ONPROC(p); - + for (;;) { + my_cred = kauth_cred_proc_ref(p); + my_new_cred = kauth_cred_setsvuidgid(my_cred, kauth_cred_getuid(my_cred), kauth_cred_getgid(my_cred)); + + if (my_new_cred == my_cred) { + kauth_cred_unref(&my_cred); + break; + } + + /* update cred on proc */ + proc_ucred_lock(p); + + if (p->p_ucred != my_cred) { + proc_ucred_unlock(p); + kauth_cred_unref(&my_new_cred); + continue; + } + + /* donate cred reference on my_new_cred to p->p_ucred */ + p->p_ucred = my_new_cred; + PROC_UPDATE_CREDS_ONPROC(p); + proc_ucred_unlock(p); + + /* drop additional reference that was taken on the previous cred */ + kauth_cred_unref(&my_cred); + + break; + } + + /* Update the process' identity version and set the security token */ - p->p_idversion++; - set_security_token(p); + p->p_idversion = OSIncrementAtomic(&nextpidversion); + + if (imgp->ip_new_thread != NULL) { + task = get_threadtask(imgp->ip_new_thread); + } else { + task = p->task; + } + set_security_token_task_internal(p, task); - return(error); + return error; } @@ -4318,20 +5134,23 @@ handle_mac_transition: * !KERN_SUCCESS Mach failure code */ static kern_return_t -create_unix_stack(vm_map_t map, load_result_t* load_result, - proc_t p) +create_unix_stack(vm_map_t map, load_result_t* load_result, + proc_t p) { - mach_vm_size_t size, prot_size; - mach_vm_offset_t addr, prot_addr; - kern_return_t kr; + mach_vm_size_t size, prot_size; + mach_vm_offset_t addr, prot_addr; + kern_return_t kr; + + mach_vm_address_t user_stack = load_result->user_stack; - mach_vm_address_t user_stack = load_result->user_stack; - proc_lock(p); p->user_stack = user_stack; + if (load_result->custom_stack) { + p->p_lflag |= P_LCUSTOM_STACK; + } proc_unlock(p); - if (!load_result->prog_allocated_stack) { + if (load_result->user_stack_alloc_size > 0) { /* * Allocate enough space for the maximum stack size we * will ever authorize and an extra page to act as @@ -4339,22 +5158,20 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, * vm_initial_limit_stack takes care of the extra guard page. * Otherwise we must allocate it ourselves. */ - - size = mach_vm_round_page(load_result->user_stack_size); - if (load_result->prog_stack_size) - size += PAGE_SIZE; + if (mach_vm_round_page_overflow(load_result->user_stack_alloc_size, &size)) { + return KERN_INVALID_ARGUMENT; + } addr = mach_vm_trunc_page(load_result->user_stack - size); - kr = mach_vm_allocate(map, &addr, size, - VM_MAKE_TAG(VM_MEMORY_STACK) | - VM_FLAGS_FIXED); + kr = mach_vm_allocate_kernel(map, &addr, size, + VM_FLAGS_FIXED, VM_MEMORY_STACK); if (kr != KERN_SUCCESS) { - /* If can't allocate at default location, try anywhere */ + // Can't allocate at default location, try anywhere addr = 0; - kr = mach_vm_allocate(map, &addr, size, - VM_MAKE_TAG(VM_MEMORY_STACK) | - VM_FLAGS_ANYWHERE); - if (kr != KERN_SUCCESS) + kr = mach_vm_allocate_kernel(map, &addr, size, + VM_FLAGS_ANYWHERE, VM_MEMORY_STACK); + if (kr != KERN_SUCCESS) { return kr; + } user_stack = addr + size; load_result->user_stack = user_stack; @@ -4364,22 +5181,29 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, proc_unlock(p); } + load_result->user_stack_alloc = addr; + /* * And prevent access to what's above the current stack * size limit for this process. */ - prot_addr = addr; - if (load_result->prog_stack_size) + if (load_result->user_stack_size == 0) { + proc_list_lock(); + load_result->user_stack_size = unix_stack_size(p); + proc_list_unlock(); + prot_size = mach_vm_trunc_page(size - load_result->user_stack_size); + } else { prot_size = PAGE_SIZE; - else - prot_size = mach_vm_trunc_page(size - unix_stack_size(p)); + } + + prot_addr = addr; kr = mach_vm_protect(map, - prot_addr, - prot_size, - FALSE, - VM_PROT_NONE); + prot_addr, + prot_size, + FALSE, + VM_PROT_NONE); if (kr != KERN_SUCCESS) { - (void) mach_vm_deallocate(map, addr, size); + (void)mach_vm_deallocate(map, addr, size); return kr; } } @@ -4400,24 +5224,24 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, * path NULL terminated path * * Returns: KERN_SUCCESS Success - * !KERN_SUCCESS See execve/mac_execve for error codes + * !KERN_SUCCESS See execve/mac_execve for error codes * * Notes: The process that is passed in is the first manufactured * process on the system, and gets here via bsd_ast() firing * for the first time. This is done to ensure that bsd_init() * has run to completion. * - * The address map of the first manufactured process is 32 bit. - * WHEN this becomes 64b, this code will fail; it needs to be - * made 64b capable. + * The address map of the first manufactured process matches the + * word width of the kernel. Once the self-exec completes, the + * initproc might be different. */ static int load_init_program_at_path(proc_t p, user_addr_t scratch_addr, const char* path) { - uint32_t argv[3]; - uint32_t argc = 0; int retval[2]; + int error; struct execve_args init_exec_args; + user_addr_t argv0 = USER_ADDR_NULL, argv1 = USER_ADDR_NULL; /* * Validate inputs and pre-conditions @@ -4426,18 +5250,17 @@ load_init_program_at_path(proc_t p, user_addr_t scratch_addr, const char* path) assert(scratch_addr); assert(path); - if (IS_64BIT_PROCESS(p)) { - panic("Init against 64b primordial proc not implemented"); - } - /* * Copy out program name. */ size_t path_length = strlen(path) + 1; - (void) copyout(path, scratch_addr, path_length); + argv0 = scratch_addr; + error = copyout(path, argv0, path_length); + if (error) { + return error; + } - argv[argc++] = (uint32_t)scratch_addr; - scratch_addr = USER_ADDR_ALIGN(scratch_addr + path_length, 16); + scratch_addr = USER_ADDR_ALIGN(scratch_addr + path_length, sizeof(user_addr_t)); /* * Put out first (and only) argument, similarly. @@ -4445,28 +5268,45 @@ load_init_program_at_path(proc_t p, user_addr_t scratch_addr, const char* path) */ if (boothowto & RB_SINGLE) { const char *init_args = "-s"; - size_t init_args_length = strlen(init_args)+1; + size_t init_args_length = strlen(init_args) + 1; - copyout(init_args, scratch_addr, init_args_length); + argv1 = scratch_addr; + error = copyout(init_args, argv1, init_args_length); + if (error) { + return error; + } - argv[argc++] = (uint32_t)scratch_addr; - scratch_addr = USER_ADDR_ALIGN(scratch_addr + init_args_length, 16); + scratch_addr = USER_ADDR_ALIGN(scratch_addr + init_args_length, sizeof(user_addr_t)); } - /* - * Null-end the argument list - */ - argv[argc] = 0; - - /* - * Copy out the argument list. - */ - (void) copyout(argv, scratch_addr, sizeof(argv)); + if (proc_is64bit(p)) { + user64_addr_t argv64bit[3] = {}; + + argv64bit[0] = argv0; + argv64bit[1] = argv1; + argv64bit[2] = USER_ADDR_NULL; + + error = copyout(argv64bit, scratch_addr, sizeof(argv64bit)); + if (error) { + return error; + } + } else { + user32_addr_t argv32bit[3] = {}; + + argv32bit[0] = (user32_addr_t)argv0; + argv32bit[1] = (user32_addr_t)argv1; + argv32bit[2] = USER_ADDR_NULL; + + error = copyout(argv32bit, scratch_addr, sizeof(argv32bit)); + if (error) { + return error; + } + } /* * Set up argument block for fake call to execve. */ - init_exec_args.fname = CAST_USER_ADDR_T(argv[0]); + init_exec_args.fname = argv0; init_exec_args.argp = scratch_addr; init_exec_args.envp = USER_ADDR_NULL; @@ -4483,7 +5323,6 @@ static const char * init_programs[] = { "/usr/local/sbin/launchd.debug", #endif #if DEVELOPMENT || DEBUG - /* Remove DEBUG conditional when is fixed */ "/usr/local/sbin/launchd.development", #endif "/sbin/launchd", @@ -4509,9 +5348,6 @@ static const char * init_programs[] = { * the kcsuffix boot-arg, setting launchdsuffix to "" or "release" * will force /sbin/launchd to be selected. * - * The DEBUG kernel will continue to check for a .development - * version until is fixed. - * * Search order by build: * * DEBUG DEVELOPMENT RELEASE PATH @@ -4526,12 +5362,14 @@ load_init_program(proc_t p) { uint32_t i; int error; - vm_offset_t scratch_addr = VM_MIN_ADDRESS; + vm_map_t map = current_map(); + mach_vm_offset_t scratch_addr = 0; + mach_vm_size_t map_page_size = vm_map_page_size(map); - (void) vm_allocate(current_map(), &scratch_addr, PAGE_SIZE, VM_FLAGS_ANYWHERE); -#if CONFIG_MEMORYSTATUS && CONFIG_JETSAM + (void) mach_vm_allocate_kernel(map, &scratch_addr, map_page_size, VM_FLAGS_ANYWHERE, VM_KERN_MEMORY_NONE); +#if CONFIG_MEMORYSTATUS (void) memorystatus_init_at_boot_snapshot(); -#endif /* CONFIG_MEMORYSTATUS && CONFIG_JETSAM */ +#endif /* CONFIG_MEMORYSTATUS */ #if DEBUG || DEVELOPMENT /* Check for boot-arg suffix first */ @@ -4539,35 +5377,43 @@ load_init_program(proc_t p) if (PE_parse_boot_argn("launchdsuffix", launchd_suffix, sizeof(launchd_suffix))) { char launchd_path[128]; boolean_t is_release_suffix = ((launchd_suffix[0] == 0) || - (strcmp(launchd_suffix, "release") == 0)); + (strcmp(launchd_suffix, "release") == 0)); if (is_release_suffix) { - error = load_init_program_at_path(p, CAST_USER_ADDR_T(scratch_addr), "/sbin/launchd"); - if (!error) + printf("load_init_program: attempting to load /sbin/launchd\n"); + error = load_init_program_at_path(p, (user_addr_t)scratch_addr, "/sbin/launchd"); + if (!error) { return; + } panic("Process 1 exec of launchd.release failed, errno %d", error); } else { strlcpy(launchd_path, "/usr/local/sbin/launchd.", sizeof(launchd_path)); strlcat(launchd_path, launchd_suffix, sizeof(launchd_path)); - /* All the error data is lost in the loop below, don't - * attempt to save it. */ - if (!load_init_program_at_path(p, CAST_USER_ADDR_T(scratch_addr), launchd_path)) { + printf("load_init_program: attempting to load %s\n", launchd_path); + error = load_init_program_at_path(p, (user_addr_t)scratch_addr, launchd_path); + if (!error) { return; + } else { + printf("load_init_program: failed loading %s: errno %d\n", launchd_path, error); } } } #endif error = ENOENT; - for (i = 0; i < sizeof(init_programs)/sizeof(init_programs[0]); i++) { - error = load_init_program_at_path(p, CAST_USER_ADDR_T(scratch_addr), init_programs[i]); - if (!error) + for (i = 0; i < sizeof(init_programs) / sizeof(init_programs[0]); i++) { + printf("load_init_program: attempting to load %s\n", init_programs[i]); + error = load_init_program_at_path(p, (user_addr_t)scratch_addr, init_programs[i]); + if (!error) { return; + } else { + printf("load_init_program: failed loading %s: errno %d\n", init_programs[i], error); + } } - panic("Process 1 exec of %s failed, errno %d", ((i == 0) ? "" : init_programs[i-1]), error); + panic("Process 1 exec of %s failed, errno %d", ((i == 0) ? "" : init_programs[i - 1]), error); } /* @@ -4589,15 +5435,17 @@ load_init_program(proc_t p) * EIO An I/O error occurred * EBADEXEC The executable is corrupt/unknown */ -static int +static int load_return_to_errno(load_return_t lrtn) { switch (lrtn) { case LOAD_SUCCESS: return 0; case LOAD_BADARCH: + case LOAD_BADARCH_X86: return EBADARCH; case LOAD_BADMACHO: + case LOAD_BADMACHO_UPX: return EBADMACHO; case LOAD_SHLIB: return ESHLIBVERS; @@ -4662,29 +5510,34 @@ static int execargs_waiters = 0; lck_mtx_t *execargs_cache_lock; static void -execargs_lock_lock(void) { +execargs_lock_lock(void) +{ lck_mtx_lock_spin(execargs_cache_lock); } static void -execargs_lock_unlock(void) { +execargs_lock_unlock(void) +{ lck_mtx_unlock(execargs_cache_lock); } static wait_result_t -execargs_lock_sleep(void) { - return(lck_mtx_sleep(execargs_cache_lock, LCK_SLEEP_DEFAULT, &execargs_free_count, THREAD_INTERRUPTIBLE)); +execargs_lock_sleep(void) +{ + return lck_mtx_sleep(execargs_cache_lock, LCK_SLEEP_DEFAULT, &execargs_free_count, THREAD_INTERRUPTIBLE); } static kern_return_t -execargs_purgeable_allocate(char **execarg_address) { - kern_return_t kr = vm_allocate(bsd_pageable_map, (vm_offset_t *)execarg_address, BSD_PAGEABLE_SIZE_PER_EXEC, VM_FLAGS_ANYWHERE | VM_FLAGS_PURGABLE); +execargs_purgeable_allocate(char **execarg_address) +{ + kern_return_t kr = vm_allocate_kernel(bsd_pageable_map, (vm_offset_t *)execarg_address, BSD_PAGEABLE_SIZE_PER_EXEC, VM_FLAGS_ANYWHERE | VM_FLAGS_PURGABLE, VM_KERN_MEMORY_NONE); assert(kr == KERN_SUCCESS); return kr; } static kern_return_t -execargs_purgeable_reference(void *execarg_address) { +execargs_purgeable_reference(void *execarg_address) +{ int state = VM_PURGABLE_NONVOLATILE; kern_return_t kr = vm_purgable_control(bsd_pageable_map, (vm_offset_t) execarg_address, VM_PURGABLE_SET_STATE, &state); @@ -4693,7 +5546,8 @@ execargs_purgeable_reference(void *execarg_address) { } static kern_return_t -execargs_purgeable_volatilize(void *execarg_address) { +execargs_purgeable_volatilize(void *execarg_address) +{ int state = VM_PURGABLE_VOLATILE | VM_PURGABLE_ORDERING_OBSOLETE; kern_return_t kr; kr = vm_purgable_control(bsd_pageable_map, (vm_offset_t) execarg_address, VM_PURGABLE_SET_STATE, &state); @@ -4704,7 +5558,8 @@ execargs_purgeable_volatilize(void *execarg_address) { } static void -execargs_wakeup_waiters(void) { +execargs_wakeup_waiters(void) +{ thread_wakeup(&execargs_free_count); } @@ -4723,7 +5578,7 @@ execargs_alloc(struct image_params *imgp) execargs_waiters--; if (res != THREAD_AWAKENED) { execargs_lock_unlock(); - return (EINTR); + return EINTR; } } @@ -4742,25 +5597,25 @@ execargs_alloc(struct image_params *imgp) assert(execargs_free_count >= 0); execargs_lock_unlock(); - + if (cache_index == -1) { kret = execargs_purgeable_allocate(&imgp->ip_strings); - } - else + } else { kret = execargs_purgeable_reference(imgp->ip_strings); + } assert(kret == KERN_SUCCESS); if (kret != KERN_SUCCESS) { - return (ENOMEM); + return ENOMEM; } /* last page used to read in file headers */ - imgp->ip_vdata = imgp->ip_strings + ( NCARGS + PAGE_SIZE ); + imgp->ip_vdata = imgp->ip_strings + (NCARGS + PAGE_SIZE); imgp->ip_strendp = imgp->ip_strings; imgp->ip_argspace = NCARGS; - imgp->ip_strspace = ( NCARGS + PAGE_SIZE ); + imgp->ip_strspace = (NCARGS + PAGE_SIZE); - return (0); + return 0; } /* @@ -4782,7 +5637,7 @@ execargs_free(struct image_params *imgp) kern_return_t kret; int i; boolean_t needs_wakeup = FALSE; - + kret = execargs_purgeable_volatilize(imgp->ip_strings); execargs_lock_lock(); @@ -4799,15 +5654,17 @@ execargs_free(struct image_params *imgp) assert(imgp->ip_strings == NULL); - if (execargs_waiters > 0) + if (execargs_waiters > 0) { needs_wakeup = TRUE; - + } + execargs_lock_unlock(); - if (needs_wakeup == TRUE) + if (needs_wakeup == TRUE) { execargs_wakeup_waiters(); + } - return ((kret == KERN_SUCCESS ? 0 : EINVAL)); + return kret == KERN_SUCCESS ? 0 : EINVAL; } static void @@ -4821,8 +5678,9 @@ exec_resettextvp(proc_t p, struct image_params *imgp) vp = imgp->ip_vp; offset = imgp->ip_arch_offset; - if (vp == NULLVP) + if (vp == NULLVP) { panic("exec_resettextvp: expected valid vp"); + } ret = vnode_ref(vp); proc_lock(p); @@ -4830,83 +5688,47 @@ exec_resettextvp(proc_t p, struct image_params *imgp) p->p_textvp = vp; p->p_textoff = offset; } else { - p->p_textvp = NULLVP; /* this is paranoia */ + p->p_textvp = NULLVP; /* this is paranoia */ p->p_textoff = 0; } proc_unlock(p); - if ( tvp != NULLVP) { + if (tvp != NULLVP) { if (vnode_getwithref(tvp) == 0) { vnode_rele(tvp); vnode_put(tvp); } - } - + } } -/* - * If the process is not signed or if it contains entitlements, we - * need to communicate through the task_access_port to taskgated. - * - * taskgated will provide a detached code signature if present, and - * will enforce any restrictions on entitlements. - */ +// Includes the 0-byte (therefore "SIZE" instead of "LEN"). +static const size_t CS_CDHASH_STRING_SIZE = CS_CDHASH_LEN * 2 + 1; -static boolean_t -taskgated_required(proc_t p, boolean_t *require_success) +static void +cdhash_to_string(char str[CS_CDHASH_STRING_SIZE], uint8_t const * const cdhash) { - size_t length; - void *blob; - int error; - - if (cs_debug > 2) - csvnode_print_debug(p->p_textvp); - - const int can_skip_taskgated = csproc_get_platform_binary(p) && !csproc_get_platform_path(p); - if (can_skip_taskgated) { - if (cs_debug) printf("taskgated not required for: %s\n", p->p_name); - *require_success = FALSE; - return FALSE; - } - - if ((p->p_csflags & CS_VALID) == 0) { - *require_success = FALSE; - return TRUE; - } + static char const nibble[] = "0123456789abcdef"; - error = cs_entitlements_blob_get(p, &blob, &length); - if (error == 0 && blob != NULL) { - /* - * fatal on the desktop when entitlements are present, - * unless we started in single-user mode - */ - if ((boothowto & RB_SINGLE) == 0) - *require_success = TRUE; - /* - * Allow initproc to run without causing taskgated to launch - */ - if (p == initproc) { - *require_success = FALSE; - return FALSE; - } - - if (cs_debug) printf("taskgated required for: %s\n", p->p_name); - - return TRUE; + /* Apparently still the safest way to get a hex representation + * of binary data. + * xnu's printf routines have %*D/%20D in theory, but "not really", see: + * confusion around %*D/%nD in printf + */ + for (int i = 0; i < CS_CDHASH_LEN; ++i) { + str[i * 2] = nibble[(cdhash[i] & 0xf0) >> 4]; + str[i * 2 + 1] = nibble[cdhash[i] & 0x0f]; } - - *require_success = FALSE; - return FALSE; + str[CS_CDHASH_STRING_SIZE - 1] = 0; } /* * __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__ - * + * * Description: Waits for the userspace daemon to respond to the request - * we made. Function declared non inline to be visible in + * we made. Function declared non inline to be visible in * stackshots and spindumps as well as debugging. */ -__attribute__((noinline)) int +__attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__(mach_port_t task_access_port, int32_t new_pid) { return find_code_signature(task_access_port, new_pid); @@ -4919,10 +5741,11 @@ check_for_signature(proc_t p, struct image_params *imgp) kern_return_t kr = KERN_FAILURE; int error = EACCES; boolean_t unexpected_failure = FALSE; - unsigned char hash[SHA1_RESULTLEN]; + struct cs_blob *csb; boolean_t require_success = FALSE; int spawn = (imgp->ip_flags & IMGPF_SPAWN); int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); + os_reason_t signature_failure_reason = OS_REASON_NULL; /* * Override inherited code signing flags with the @@ -4934,40 +5757,59 @@ check_for_signature(proc_t p, struct image_params *imgp) proc_unlock(p); /* Set the switch_protect flag on the map */ - if(p->p_csflags & (CS_HARD|CS_KILL)) { + if (p->p_csflags & (CS_HARD | CS_KILL)) { vm_map_switch_protect(get_task_map(p->task), TRUE); } - + /* * image activation may be failed due to policy * which is unexpected but security framework does not * approve of exec, kill and return immediately. */ if (imgp->ip_mac_return != 0) { + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SECURITY_POLICY, 0, 0); + signature_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_SECURITY_POLICY); error = imgp->ip_mac_return; unexpected_failure = TRUE; goto done; } - /* check if callout to taskgated is needed */ - if (!taskgated_required(p, &require_success)) { + if (imgp->ip_cs_error != OS_REASON_NULL) { + signature_failure_reason = imgp->ip_cs_error; + imgp->ip_cs_error = OS_REASON_NULL; + error = EACCES; + goto done; + } + + /* If the code signature came through the image activation path, we skip the + * taskgated / externally attached path. */ + if (imgp->ip_csflags & CS_SIGNED) { error = 0; goto done; } + /* The rest of the code is for signatures that either already have been externally + * attached (likely, but not necessarily by a previous run through the taskgated + * path), or that will now be attached by taskgated. */ + kr = task_get_task_access_port(p->task, &port); if (KERN_SUCCESS != kr || !IPC_PORT_VALID(port)) { error = 0; - if (require_success) + if (require_success) { + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASK_ACCESS_PORT, 0, 0); + signature_failure_reason = os_reason_create(OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASK_ACCESS_PORT); error = EACCES; + } goto done; } /* * taskgated returns KERN_SUCCESS if it has completed its work - * and the exec should continue, KERN_FAILURE if the exec should - * fail, or it may error out with different error code in an - * event of mig failure (e.g. process was signalled during the + * and the exec should continue, KERN_FAILURE if the exec should + * fail, or it may error out with different error code in an + * event of mig failure (e.g. process was signalled during the * rpc call, taskgated died, mig server died etc.). */ @@ -4978,38 +5820,93 @@ check_for_signature(proc_t p, struct image_params *imgp) break; case KERN_FAILURE: error = EACCES; + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASKGATED_INVALID_SIG, 0, 0); + signature_failure_reason = os_reason_create(OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASKGATED_INVALID_SIG); goto done; default: error = EACCES; + + KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_TASKGATED_OTHER, 0, 0); + signature_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_TASKGATED_OTHER); unexpected_failure = TRUE; goto done; } /* Only do this if exec_resettextvp() did not fail */ if (p->p_textvp != NULLVP) { - /* - * If there's a new code directory, mark this process - * as signed. - */ - if (0 == ubc_cs_getcdhash(p->p_textvp, p->p_textoff, hash)) { - proc_lock(p); - p->p_csflags |= CS_VALID; - proc_unlock(p); + csb = ubc_cs_blob_get(p->p_textvp, -1, p->p_textoff); + + if (csb != NULL) { + /* As the enforcement we can do here is very limited, we only allow things that + * are the only reason why this code path still exists: + * Adhoc signed non-platform binaries without special cs_flags and without any + * entitlements (unrestricted ones still pass AMFI). */ + if ( + /* Revalidate the blob if necessary through bumped generation count. */ + (ubc_cs_generation_check(p->p_textvp) == 0 || + ubc_cs_blob_revalidate(p->p_textvp, csb, imgp, 0) == 0) && + /* Only CS_ADHOC, no CS_KILL, CS_HARD etc. */ + (csb->csb_flags & CS_ALLOWED_MACHO) == CS_ADHOC && + /* If it has a CMS blob, it's not adhoc. The CS_ADHOC flag can lie. */ + csblob_find_blob_bytes((const uint8_t *)csb->csb_mem_kaddr, csb->csb_mem_size, + CSSLOT_SIGNATURESLOT, + CSMAGIC_BLOBWRAPPER) == NULL && + /* It could still be in a trust cache (unlikely with CS_ADHOC), or a magic path. */ + csb->csb_platform_binary == 0 && + /* No entitlements, not even unrestricted ones. */ + csb->csb_entitlements_blob == NULL) { + proc_lock(p); + p->p_csflags |= CS_SIGNED | CS_VALID; + proc_unlock(p); + } else { + uint8_t cdhash[CS_CDHASH_LEN]; + char cdhash_string[CS_CDHASH_STRING_SIZE]; + proc_getcdhash(p, cdhash); + cdhash_to_string(cdhash_string, cdhash); + printf("ignoring detached code signature on '%s' with cdhash '%s' " + "because it is invalid, or not a simple adhoc signature.\n", + p->p_name, cdhash_string); + } } } done: - if (0 != error) { - if (!unexpected_failure) + if (0 == error) { + /* The process's code signature related properties are + * fully set up, so this is an opportune moment to log + * platform binary execution, if desired. */ + if (platform_exec_logging != 0 && csproc_get_platform_binary(p)) { + uint8_t cdhash[CS_CDHASH_LEN]; + char cdhash_string[CS_CDHASH_STRING_SIZE]; + proc_getcdhash(p, cdhash); + cdhash_to_string(cdhash_string, cdhash); + + os_log(peLog, "CS Platform Exec Logging: Executing platform signed binary " + "'%s' with cdhash %s\n", p->p_name, cdhash_string); + } + } else { + if (!unexpected_failure) { p->p_csflags |= CS_KILLED; + } /* make very sure execution fails */ if (vfexec || spawn) { - psignal_vfork(p, p->task, imgp->ip_new_thread, SIGKILL); + assert(signature_failure_reason != OS_REASON_NULL); + psignal_vfork_with_reason(p, p->task, imgp->ip_new_thread, + SIGKILL, signature_failure_reason); + signature_failure_reason = OS_REASON_NULL; error = 0; } else { - psignal(p, SIGKILL); + assert(signature_failure_reason != OS_REASON_NULL); + psignal_with_reason(p, SIGKILL, signature_failure_reason); + signature_failure_reason = OS_REASON_NULL; } } + + /* If we hit this, we likely would have leaked an exit reason */ + assert(signature_failure_reason == OS_REASON_NULL); return error; } @@ -5023,7 +5920,8 @@ done: * in the process' page tables, we prefault some pages if * possible. Errors are non-fatal. */ -static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, load_result_t *load_result) +static void +exec_prefault_data(proc_t p __unused, struct image_params *imgp, load_result_t *load_result) { int ret; size_t expected_all_image_infos_size; @@ -5032,13 +5930,13 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa * Prefault executable or dyld entry point. */ vm_fault(current_map(), - vm_map_trunc_page(load_result->entry_point, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, - THREAD_UNINT, NULL, 0); - - if (imgp->ip_flags & IMGPF_IS_64BIT) { + vm_map_trunc_page(load_result->entry_point, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); + + if (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) { expected_all_image_infos_size = sizeof(struct user64_dyld_all_image_infos); } else { expected_all_image_infos_size = sizeof(struct user32_dyld_all_image_infos); @@ -5046,11 +5944,11 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa /* Decode dyld anchor structure from */ if (load_result->dynlinker && - load_result->all_image_info_addr && - load_result->all_image_info_size >= expected_all_image_infos_size) { + load_result->all_image_info_addr && + load_result->all_image_info_size >= expected_all_image_infos_size) { union { - struct user64_dyld_all_image_infos infos64; - struct user32_dyld_all_image_infos infos32; + struct user64_dyld_all_image_infos infos64; + struct user32_dyld_all_image_infos infos32; } all_image_infos; /* @@ -5058,33 +5956,32 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa * and recovery path. */ vm_fault(current_map(), - vm_map_trunc_page(load_result->all_image_info_addr, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(load_result->all_image_info_addr, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); if ((load_result->all_image_info_addr & PAGE_MASK) + expected_all_image_infos_size > PAGE_SIZE) { /* all_image_infos straddles a page */ vm_fault(current_map(), - vm_map_trunc_page(load_result->all_image_info_addr + expected_all_image_infos_size - 1, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(load_result->all_image_info_addr + expected_all_image_infos_size - 1, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); } ret = copyin(load_result->all_image_info_addr, - &all_image_infos, - expected_all_image_infos_size); - if (ret == 0 && all_image_infos.infos32.version >= 9) { - + &all_image_infos, + expected_all_image_infos_size); + if (ret == 0 && all_image_infos.infos32.version >= DYLD_ALL_IMAGE_INFOS_ADDRESS_MINIMUM_VERSION) { user_addr_t notification_address; user_addr_t dyld_image_address; user_addr_t dyld_version_address; user_addr_t dyld_all_image_infos_address; user_addr_t dyld_slide_amount; - if (imgp->ip_flags & IMGPF_IS_64BIT) { + if (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) { notification_address = all_image_infos.infos64.notification; dyld_image_address = all_image_infos.infos64.dyldImageLoadAddress; dyld_version_address = all_image_infos.infos64.dyldVersion; @@ -5113,38 +6010,38 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa #if 0 kprintf("exec_prefault: 0x%016llx 0x%08x 0x%016llx 0x%016llx 0x%016llx 0x%016llx\n", - (uint64_t)load_result->all_image_info_addr, - all_image_infos.infos32.version, - (uint64_t)notification_address, - (uint64_t)dyld_image_address, - (uint64_t)dyld_version_address, - (uint64_t)dyld_all_image_infos_address); + (uint64_t)load_result->all_image_info_addr, + all_image_infos.infos32.version, + (uint64_t)notification_address, + (uint64_t)dyld_image_address, + (uint64_t)dyld_version_address, + (uint64_t)dyld_all_image_infos_address); #endif vm_fault(current_map(), - vm_map_trunc_page(notification_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(notification_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_image_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_image_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_version_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_version_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_all_image_infos_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_all_image_infos_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); } } }